Applications of Close-Range Terrestrial 3D Photogrammetry

to Improve Safety in Underground Stone Mines

Richard Edwin Bishop

Thesis submitted to the Faculty of the

Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of

Master of Science
in
Mining Engineering

Nino S. Ripepi, Chair

Cheng Chen
Michael E. Karmis

May 22, 2020

Blacksburg, VA

Keywords: photogrammetry, stereophotogrammetry, underground, mining, mine safety

Copyright 2020, Richard E. Bishop

 Applications of Close-Range Terrestrial 3D Photogrammetry
to Improve Safety in Underground Stone Mines
Richard Edwin Bishop

ACADEMIC ABSTRACT

The underground limestone mining industry is a small, but growing segment of the

U.S. crushed stone industry. However, its fatality rate has been amongst the highest of the

mining sector in recent years due to ground control issues related to ground collapses. It is

therefore important to improve the engineering design, monitoring and visualization of

ground control by utilizing new technologies that can help an underground limestone

company maintain a safe and productive operation.

Photogrammetry and laser scanning are remote sensing technologies that are useful

tools for collecting three-dimensional spatial data with high levels of precision for many

types of mining applications. Due to the reality of budget constraints for many underground

stone mining operations, this research concentrates on photogrammetry as a more

accessible technology for the average operation. Despite the challenging lighting

conditions and size of underground limestone mines that has previous hindered

photogrammetric surveys in these environments, over 13,000 photographic images were

taken over a 3-year period in active mines to compile these models. This research

summarizes that work and highlights the many applications of terrestrial close-range

photogrammetry, including practical methodologies for implementing the techniques in

working operations to better visualize hazards and pragmatic approaches for geotechnical

analysis, improved engineering design and monitoring.

 Applications of Close-Range Terrestrial 3D Photogrammetry
to Improve Safety in Underground Stone Mines
Richard Edwin Bishop

GENERAL AUDIENCE ABSTRACT

The underground limestone mining industry is a small, but growing segment of the

U.S. crushed stone industry. However, its fatality rate has been amongst the highest of the

mining sector in recent years due to ground control issues related to ground collapses. It is

therefore important to improve the engineering design, monitoring and visualization of

ground control by utilizing new technologies that can help maintain safe and productive

underground stone operations. Photogrammetry and laser scanning are remote sensing

technologies that are useful tools for collecting three-dimensional spatial data with high

levels of precision for many different mining applications. Due to the reality of budget

constraints for many mining operations, this research concentrates on photogrammetry as

a more accessible technology for the average operation, despite the challenging lighting

conditions and expansive size of underground limestone mines that has previous hindered

photogrammetric surveys in these environments. This research focuses on the applications

of photogrammetry in underground stone mines and practical methodologies for

implementing the techniques in working operations to better visualize hazards for

improved engineering design and infrastructure management.

Dedication

This work is dedicated to my late grandfather, Richard M. Bishop, a coal miner and

to other miners, past, present and future who dedicate their lives to their families, the

industry and to society by bravely providing the raw materials that bring quality to our

lives. May this work inspire improvements and new ideas for the safety and health of mine

workers.

iv
Acknowledgments

Thank you to NIOSH for the funding and support to complete this research under

contract number 200-2016-91300, as well as the ongoing professional encouragement from

Brent Slaker and Michael Murphy to develop new ideas and methodologies to improve the

safety of our industry. Thank you Maptek for providing access and support to the

PointStudio software that was utilized during this research. Thank you Mark Luxbacher

and Nathan Bench accommodating mine access and contributing ideas to the project.

Many thanks to my research team, Jon Baggett, Juan Monsalve and Aman Soni, for

your help on this project. Your effort on this project has been helpful and a pleasure to

collaborate. I would also like to thank Alex Scheck and Austin Fox for your assistance

during the course of this research. I appreciate all of the help building equipment and

assistance underground in mines collecting data.

Thank you to those who served as mentors throughout my industry career: Bill

Walker, David Stein, Stephen Walker, Michael Curran, Karen Mikkola, Brian Ball, Shawn

Conaway and many others. You each motivated me to continuously improve myself

professionally, for which I am most thankful.

Thank you to my professors over my academic career, Christopher Haycocks, Roe-

Hoan Yoon, Michael Karmis, Greg Adel, Jerry Luttrell, Mario Karfakis, Erik Westman,

Kray Luxbacher, Emily Sarver, and many others. A special thanks to my graduate advisor,

Nino Ripepi, for your support and encouragement to complete my graduate studies in

mining engineering. My hope is to be able to give back to our industry by educating and

inspiring the next generation of mining engineers as you all have done for me.

v
Preface

This thesis is composed of five chapters in a traditional format, except for Chapter

5 which was adapted from a published conference proceeding. This chapter titled “A

Comparison of Laser Scanning and Photogrammetry in an Underground Limestone Mine”

was presented and published in conference proceedings as part of the 2019 Annual Society

of Mining, Metallurgy & Exploration (SME) Meeting in Denver, Colorado. It was co-

authored by Juan Monsalve, Jon Baggett, Aman Soni and Dr. Nino Ripepi. The conclusions

for the chapter are merged into the final conclusions in Chapter 6. Permission for release

from SME can be seen in Appendix D.

vi
Table of Contents

Chapter 1 Introduction ................................................................................. 1

Chapter 2 Literature Review ........................................................................ 6
Practice Review .............................................................................................6
Monitoring and Preventing Fall of Ground .................................................10
Photogrammetry ..........................................................................................15
Camera Terminology and Settings ..............................................................17
Software.......................................................................................................21
Risk Management ........................................................................................24
Chapter 3 Equipment Design and Methodology ......................................26
Lighting Considerations ..............................................................................26
Lighting Test ...............................................................................................28
Lens Field of View Test ..............................................................................31
Camera Setting Test ....................................................................................34
Lens Correction ...........................................................................................35
Monocular Photogrammetry........................................................................37
Stereophotogrammetry ................................................................................39
Computational Processing Time .................................................................44
Photogrammetry Workflow.........................................................................48
Chapter 4 Applications ...............................................................................51
Mine Risk Databases ...................................................................................51
Mine Headings ............................................................................................52
Change Detection ........................................................................................53
Karst Mapping .............................................................................................55
Ground Control............................................................................................56
As-Built Underground Mine Construction Surveys ....................................58
Pillar Monitoring .........................................................................................59
Geotechnical Analysis .................................................................................61
Highwall Inspections ...................................................................................62
Smartphone Photogrammetry ......................................................................63

vii
Chapter 5 Comparison of Laser Scanning and Photogrammetry .......... 64
Introduction .................................................................................................65
Equipment and Methodology ......................................................................68
Data Processing & Analysis ........................................................................72
Advantages & Limitations...........................................................................76
Chapter 6 Conclusions and Future Work .................................................79
Conclusions .................................................................................................79
Future Work: Virtual & Augmented Reality Applications .........................82
Future Work: Autonomous Drones .............................................................83
Bibliography .................................................................................................86
Appendix A: Photogrammetry Model Summaries...................................88
Appendix B: Camera Setting Test Data ....................................................90
Appendix C: Photogrammetry Image Sets ...............................................91
Pillar A (Full-frame DSLR 35mm) .............................................................91
Pillar A (Full-frame DSLR 35mm) .............................................................99
Pillar B (UAV) ..........................................................................................101
Karst (Full-frame DSLR 35mm) ...............................................................109
Steel Sets (Full-frame DSLR 35mm) ........................................................113
Pillar Face (UAV) .....................................................................................119
Building (UAV) .........................................................................................121
Heading 1 (Full-frame DSLR 35mm) .......................................................143
Heading 2 (Full-frame DSLR 35mm) .......................................................147
Heading 3 (Full-frame DSLR 35mm) .......................................................153
Heading 4 (Full-frame DSLR 35mm) .......................................................157
Mine Drift with Karst (Full-frame DSLR 35mm) .....................................161
Mine Rib (iPhone 6) ..................................................................................179
Appendix D: Permission for Release .......................................................183
SME 2019 Preprint (19-095) .....................................................................183

viii
List of Figures
Figure 1-1: Active Underground Stone Mining Operations in the USA ............................ 1
Figure 1-2: Permitting risk between surface and underground quarries ............................. 3
Figure 1-3: Number and % fatalities by accident class at underground mines ................... 4
Figure 1-4: Expansive Size of an Underground Limestone Mine ...................................... 5
Figure 2-1: Encountering Karst Underground .................................................................... 7
Figure 2-2: Karst Trend Through Three Parallel Mine Headings....................................... 8
Figure 2-3: Parallel Headings with Grout Plugging Karst Feature ..................................... 8
Figure 2-4: Steel Bands and Mesh for Ground Control over Karst Void ........................... 9
Figure 2-5: Fall of Ground in an Underground Limestone Mine ..................................... 10
Figure 2-6: 3D Point Cloud of an Underground Limestone Mine .................................... 12
Figure 2-7: Laser Scans of an Underground Limestone Mine Level................................ 12
Figure 2-8: Traditional Underground Mine Surveying with a Theodolite ....................... 14
Figure 2-9: Using Photogrammetry to Identify Points in 3D Space ................................. 16
Figure 2-10: The Exposure Triangle ................................................................................. 17
Figure 2-11: Example Histograms Indicating Image Exposure........................................ 20
Figure 2-12: Depth of field of a lens ................................................................................. 21
Figure 2-13: Roof Fall Risk Map ...................................................................................... 25
Figure 3-1: Example of Image Underexposure due to Spot Lighting ............................... 27
Figure 3-2: Initial Underground Photogrammetry Testing ............................................... 28
Figure 3-3: 30-watt LED’s with Battery Mounted to 24” Steel Bar ................................. 29
Figure 3-4: Lumen Meter Used for Measuring Light Intensity ........................................ 29
Figure 3-5: Brightness of Lights Tested at Various Distances ......................................... 30
Figure 3-6: Pillar Photographed for Lens and ISO Tests .................................................. 31
Figure 3-7: Determining the Optimal Lens Focal Length ................................................ 32
Figure 3-8: Impact of Full-Frame vs Crop Sensor Camera .............................................. 33
Figure 3-9: ISO vs Exposure at Varying Aperture Settings ............................................. 35
Figure 3-10: Lens Correction by Grid Analysis ............................................................... 35
Figure 3-11: Radial & Decentering Distortion (35mm lens) ............................................ 36
Figure 3-12: DSLR Photogrammetry Using a Monopod .................................................. 38

ix
Figure 3-13: Stereoscopic Camera Storage Protection ..................................................... 39
Figure 3-14: Wireless Remote Shutter Control for Stereopair Photogrammetry ............. 40
Figure 3-15: Laser Sight with Hot Shoe Adapter to Assist Camera Alignment ............... 41
Figure 3-16: Stereophotogrammetry Rig .......................................................................... 42
Figure 3-17: Photogrammetry Model of a Building ......................................................... 45
Figure 3-18: Location of Images Used for Photogrammetry Model ................................ 45
Figure 3-19: Benchmark Processing Time for Building Model ....................................... 47
Figure 3-20: Photogrammetric Survey Around a Typical Mine Pillar ............................. 48
Figure 3-21: Photogrammetry Workflow ......................................................................... 49
Figure 4-1: Mine Heading Photogrammetry Model Screenshots ..................................... 52
Figure 4-2: Change Detection between Photogrammetry Models After Hand Scaling.... 53
Figure 4-3: Photogrammetry Model Analyzed for Blast Fragmentation .......................... 54
Figure 4-4: Photogrammetry Survey of an Open Karst Void ........................................... 55
Figure 4-5: Karst Model.................................................................................................... 56
Figure 4-6: Drone Photogrammetry Survey of a Pillar with Reinforcement .................... 57
Figure 4-7: Drone Photogrammetry Model of Rock Bolt and Mesh Reinforced Pillar.... 57
Figure 4-8: Photograph of an Underground Construction Project .................................... 58
Figure 4-9: As-Built Photogrammetry Survey of Steel Set Ground Support ................... 59
Figure 4-10: Fully Modelled Mine Pillar Using Full Frame DSLR ................................. 60
Figure 4-11: Fully Modelled Mine Pillar via Drone-based Photogrammetry................... 60
Figure 4-12: Geotechnical Mapping of a Point Cloud Obtained from Photogrammetry . 61
Figure 4-13: Mine Entrance Highwall Modelled via Drone-based Photogrammetry....... 62
Figure 4-14: Underground Mine Photogrammetry from a Smartphone ........................... 63
Figure 5-1: 3D point cloud of the face of subject limestone pillar from laser scanning... 67
Figure 5-2: 3D rendering of subject pillar using photogrammetry ................................... 68
Figure 5-3: Effect of focal length on field of view ........................................................... 69
Figure 5-4: Photogrammetry survey of mine pillar .......................................................... 70
Figure 5-5: Survey overview map..................................................................................... 71
Figure 5-6: Laser scanning of a limestone mine pillar ..................................................... 72
Figure 5-7: Point Clouds – Laser Scan vs Photogrammetry ............................................. 74
Figure 5-8: Distance error analysis ................................................................................... 75

x
Figure 5-9: Merged laser scan + photogrammetry............................................................ 78
Figure 6-1: Virtual Reality for 3D Visualization of Underground Mine Models ............. 83
Figure 6-2: Augmented Reality of a Mine Pillar .............................................................. 83
Figure 6-3: Drone with LED’s Mapping a 100-foot Tall Entry with Photogrammetry.... 85

xi
List of Tables

Table 1: Lens Correction Profile Created for 35mm Lens ............................................... 36

Table 2: Monocular Photogrammetry Equipment List ..................................................... 37
Table 3: Stereoscopic Photogrammetry Equipment List .................................................. 43
Table 4: Exposure Time at Various Aperture and ISO Settings ....................................... 90

xii
Chapter 1

Introduction

The United States crushed stone market is a ~$18.7 billion per year industry

representing over 3,000 mining operations across all 50 states. In 2019 alone, 1.53 billion

tons of crushed stone was produced with 72% used predominately for road construction

and repairs, but the remaining production is also important for the cement, lime, chemical

and agricultural industries. Among the domestic producers of crushed stone in this country,

over 69% source their material from limestone and dolomite deposits, 15% from granite,

and the balance from traprock, sandstone, and other miscellaneous stones (USGS, 2020).

The 114 active underground stone operations in the U.S. are primarily concentrated in the

eastern and central portion of the country as seen in Figure 1-1 (NIOSH, 2018).

Figure 1-1: Active Underground Stone Mining Operations in the USA (NIOSH, 2018)

1
 Despite annual increases in crushed stone production in the U.S., the overall

number of crushed stone mines in the U.S. is decreasing in favor for larger operations with

greater economies of scale and expanding production capacities. As cities and metropolitan

areas compete for land use with surface quarries and new permits being more difficult to

acquire, underground limestone mines in the U.S. have become more common and the

number is expected to increase (USGS, 2012). With less experienced underground miners

relative to experienced surface miners, there is typically there is a need for more education

and training and increased engineering controls to keep workers safe in large opening

underground operations. With underground limestone mining having the highest incident

rate for underground mining in the United States, it is imperative to better understand the

nature of the hazards and take steps toward mitigating these risks.

Although production of crushed stone is currently dominated by surface mines,

increased numbers of crushed stone producers have moved to underground mining because

of several challenges. These challenges include increasing regulatory pressure with respect

to environmental impacts, land development around existing surface quarries impeding

expansion and difficulties in gaining permits to operate above ground relative to

underground. Urban sprawl can cut off otherwise mineable and economical resources, also

called resource sterilization, due to the competing land use against surface mining by

expanding land development which also makes the permitting process more difficult for

surface mines (USGS, 2012). Figure 1-2 below indicates the greater probability of failure

to permit surface quarries compared to underground mining which also helps explain why

more stone mining operations are considering going underground (Haycocks, 1993).

2
 Figure 1-2: Permitting risk between surface and underground quarries (Haycocks, 1993)

Despite underground mines being easier to permit however, underground mining

requires a different technical skillset, and poses great challenges for production that must

be both safe and economical. Therefore, there are still many risks that an operation must

mitigate in order to produce limestone from an underground mine. Hazards such as water

inflow from geologic features such as karsts, faults and structural weaknesses can be

difficult to assess in a production environment (NIOSH, 1998).

According to data provided by the Mine Safety and Health Administration

(MSHA), the underground stone mining industry has had a fatality rate amongst the highest

in the mining industry in recent years (MSHA, 2016). Over the past 35 years, MSHA data

has indicated that 40% of underground mining fatalities in stone mining operations were

caused by ground control issues related to ground collapses as seen in Figure 1-3. NIOSH

research has affirmed that underground stone mining has had a high fatality concentration

due to ground failures, with ~92% (12 of the 13 fatalities in their study) due to falls of face,

rib, pillar, side, highwall, roof or back (NIOSH, 1998).

3
 Figure 1-3: Number and % fatalities by accident class at underground mines, 1983 – 2018
(MSHA data)

As illustrated in Figure 1-4, the expansive size and roof height of many

underground stone mines can make roof fall hazards exceptionally dangerous and

incredibly difficult to see from the ground. Some of the challenges in underground

limestone mining involve geological hazards that are unique to the rest of the industry. The

consequences of failure due to improperly designed underground mining ground support

systems, such as from a roof failure, are not only life threatening for the miners themselves,

but can easily shut down an entire mining operation. This also impacts the mining company

by damaging its reputation and can strip it of its social license to operate and build new

mines. The consequential risk to miners from fall of ground remains the highest fatality

exposure risk in underground stone mines.

4
Figure 1-4: Expansive Size of an Underground Limestone Mine

5
Chapter 2

Literature Review

Practice Review

More than 30 visits were made to 3 different underground limestone operations

while conducting this research project. During those visits, discussions were made with

mine management, engineers and safety personnel with regards to ground control practices

in their respective operations. In most of those operations, consultants and researchers have

conducted previous studies utilizing numerical modeling, laser scanning, and

photogrammetry to address ground control monitoring, issues and pillar design. It was

observed in each operation that remote sensing for ground control monitoring and

numerical modeling are not currently done at the site level, but the previous studies all

showed promising application.

The operations use various forms of roof control such as rock bolting, mesh,

grouting, shotcrete, steel straps and steel sets to strengthen and shield the roof and rib in an

effort to prevent ground falls on an as-needed basis. Mine pillars are designed empirically

and are oversized with a higher factor of safety when compared with the recommendations

made by NIOSH (Esterhuizen et al., 2005). Miners maintain safety in ground control by

regular scaling of loose rock from the mine face, rib and roof.

One of the biggest hazards that underground limestone mines deal with on a regular

basis is water. As limestone is associated with karst terrain, the operation may encounter

6
karstic voids filled with mud and/or water and loose rock during the mining process. An

area of an underground limestone mine intersecting a karst feature is seen in Figure 2-1.

Figure 2-1: Encountering Karst Underground

Some of the operations currently use probe drilling (i.e. holes drilled in advance of

blasting to detect water at the mine face) in an effort to detect karst ahead of mining. When

encountered, the miners pump in grout to try and seal off the voids and continue mining.

As observed from a combination of laser scans from an underground limestone mine level,

a karstic trend along three parallel headings together with examples of grout plugging karst

in all three headings can be seen in Figure 2-2 and Figure 2-3, respectively. All three

7
 Figure 2-2: Karst Trend Through Three Parallel Mine Headings

Figure 2-3: Parallel Headings with Grout Plugging Karst Feature (Left to Right on Above Map)

parallel headings in this real-world application intersected a karst void and were sealed

with grout. If the karst area is determined to be too large to plug with grout and difficult to

mine through, an effort is made to modify the mine plan to go around the area. When

mining through a large karst, additional capital is required to install protective ground

control measures. An example of a protective mesh, rock bolt and steel strap installation

used to protect mine workers from mud and rock falls from a karst opening in an

underground limestone mine can be seen in Figure 2-4. While some inference can be made

from the trend observed in Figure 2-2 as to the likelihood of a karst and the size of the

8
 Figure 2-4: Steel Bands and Mesh for Ground Control over Karst Void

karst to be expected in a parallel heading, most mine development lacks geophysical data

to make informed decisions about the ground conditions in the mine face ahead.

Management from one of the mining companies thought that ground penetrating radar

(GPR) would be very beneficial to an earlier detection of karst voids over the current

method of probe drilling. If a mine had more awareness of karst voids and water issues in

advance of the active mine face, it would be better positioned to make appropriate changes

in the mine plan to avoid the area. Previous research by Baggett et al. have validated the

applicability of GPR technology for detecting karst voids in underground limestone mines.

Ground penetrating radar would also be helpful to the operation as a method to investigate

9
pillars for karst, which reduce pillar strength, and help the engineers design pillar

reinforcement accordingly before potential problems occur (Baggett, 2019).

Monitoring and Preventing Fall of Ground

Mine geology must be considered to improve the safety of each underground

limestone mining operation. Discontinuities can weaken the rock mass and allow for water

to move through limestone formations and form karsts, which are common in limestone

deposits (Kehew, 1988). Structural weaknesses along karstic interfaces and geologic

discontinuities have the potential to give way to massive blocks weighing multiple tons,

the scale of which is evident by Figure 2-5.

Figure 2-5: Fall of Ground in an Underground Limestone Mine

10
 Previous research in the area of remote imaging and sensing, including ground

penetrating radar, terrestrial laser scanning and photogrammetry, has shown benefits for

improving the safety of underground stone mining operations (Baggett, 2019). Each useful

for monitoring, characterizing and visualizing certain geological features and spatial

relationships from a distance, remote sensing techniques have the potential to help identify

and measure risks in underground limestone mines.

Terrestrial laser scanners offer a number of benefits suitable for three-dimensional

mapping of underground limestone mines. Firstly, laser scanners actively send laser light

pulses from the scanner to the surrounding surfaces without the need for any additional

light source. The laser scanner measures the time for each light pulse to reflect and return

from the surface to the scanner in order to calculate its three-dimensional position. Laser

scanners such as the Faro Focus 3D tested in this research are capable of measuring

millions of 3D points of the surrounding scene in a matter of minutes. The operator defines

the required scan parameters, pushes a button on the scanner to start, and within a few

minutes a point cloud of the scene is captured. A 3D point cloud captured within a room

and pillar limestone mine can be seen in Figure 2-6.

Laser scanning has been shown to be an effective means for three-dimensional

mapping of underground limestone workings and characterizing geologic features

important to ground control (Monsalve, Baggett, Bishop, & Ripepi, 2019). Another

advantage of laser scanning for mapping in underground limestone mines is the ability to

cover large areas. Without a GPS-signal underground, a laser scanner is unable to reference

its location, however it is able to determine orientation based upon an internal compass

(Faro, 2013). Figure 2-7 shows an example of an entire level in an underground limestone

11
mine 3D mapped with 57 adjoined stations positioned together with the Maptek

PointStudio software.

Figure 2-6: 3D Point Cloud of an Underground Limestone Mine Obtained by Laser Scanning

Figure 2-7: Laser Scans of an Underground Limestone Mine Level (Plan & Isometric View)

12
 One downside of laser scanning equipment is the initial cost, which is easily several

tens of thousands of dollars. Not all mining operations have the budget to invest in new

equipment such as laser scanners. Most operations rely upon traditional underground mine

surveying with theodolites, though many have upgraded to digital electronic theodolites,

similar to the one seen in Figure 2-8. On the other hand, many underground limestone

operations already have access to and utilize digital cameras for documenting safety related

incidents that could also be used for applying photogrammetry as a supplementary

mapping, planning and visualization tool in the engineering and management workflow.

13
Figure 2-8: Traditional Underground Mine Surveying with a Digital Electronic Theodolite

14
Photogrammetry

Photogrammetry is defined as a method of obtaining reliable images of an object

that can be used for precise metric calculations of the subject as well as interpretive

identification of features without physical contact. The practice has been used since the

dawn of photography when in 1840 photographs were demonstrated in France as a tool that

could be used for topographic surveying (Wolf, 1974). The applications have increased

over the last 180 years, from wide usage in aerial mapping, to the creation of photorealistic

3D models used in many video games (Lachambre, Lagarde, & Jover, 2017).

Photogrammetry is a readily accessible means of technology with the advent of

digital cameras and software packages to handle the computational requirements to make

the analysis work. However, it is not without unique challenges in an underground mine

environment. Photography requires adequate light to record photographic data, and while

the low light environment encountered in an underground mine can be offset with a camera

flash and supplementary lighting, dust and the large cavernous opening in underground

limestone can make it difficult to get sufficient surveys for the inexperienced operator

(Slaker, 2017).

In mining, photogrammetry is more frequently used in surface operations for it is

a technology noted to help evaluate minuscule deformations in rock structure. It is a

technology that can help characterize rock mass discontinuities and joint sets, as well as

help monitor ground control. Photogrammetry derives the geometric properties of an object

or scene from one or more photographs, which can be used to reveal the deformational

behavior of objects underground over time, including such objects as underground

limestone pillars.

15
 Photogrammetry uses two-dimensional photographic imaging to make quantitative

(metric) and qualitative (interpretative) models of real-world objects (Wolf, 1974). With

just one photograph, light hitting a camera pixel could originate from any point along the

ray from the pixel to the perspective center. By adding another image taken from a different

location of an overlapping scene, the three-dimensional location of the point where the

light originated can be determined as illustrated in Figure 2-9 below.

Figure 2-9: Using Photogrammetry to Identify Points in 3D Space

16
Camera Terminology and Settings

Many factors need to be taken into consideration when conducting a

photogrammetry survey underground. Photogrammetry requires more user experience than

laser scanning as the process is far less automated. The quality of the photogrammetry

survey depends heavily on the experience level of the camera operator (Lanmar Services,

2014).

The operator needs to be familiar with several photographic basics to ensure good

results, including shutter speed, aperture and ISO sensitivity. The relationship between

these settings is best visualized with the Exposure Triangle, as seen in Figure 2-10 below,

where the shutter speed, Tv, aperture, Av, and ISO sensitivity are all related and influence

the exposure.

Figure 2-10: The Exposure Triangle (adapted from Peterson, 2016)

17
Aperture

Aperture, Av, is the feature of a lens that lets in light and is represented by a ratio

referring to the relative size of the opening, also called the f-stop or the f-number. The f-

stop is equal to the focal length divided by the lens diameter. More light is able to reach

the sensor when the hole is larger and less light when the hole is smaller. Aperture priority

mode on a digital camera is a setting that allows for the photographer to keep the aperture

of the lens constant, while the camera processor determines the best combination of the

other two components of the exposure triangle, shutter speed and ISO sensitivity, to obtain

a balance exposure. It is usually represented by an “A” icon on a Nikon DSLR’s setting

wheel or “Av” on a Canon DSLR.

Shutter Speed

Shutter speed, measured in seconds and fractions of a second, controls how long a

shutter is open allowing light to hit the camera’s sensor, called the time value Tv. The

shorter (faster) the shutter speed, the less light is able to expose the sensor. The longer

(slower) the shutter speed, the more light is able to expose the sensor. Shutter priority mode

on a digital camera functions similarly to aperture priority mode, except that the shutter

speed is pre-selected while the camera processor adjusts the lens aperture and ISO

sensitivity to balance the image exposure. Shutter priority mode is typically represented by

the “Tv” icon on a Canon DLSR or by a “S” icon on a Nikon DSLR.

ISO Sensitivity

A camera sensor’s sensitivity is controlled by the ISO setting. When a camera’s

ISO sensitivity is set to a high value, it needs less light for a proper exposure, and needs

additional light to achieve the same exposure with a low ISO setting. Increasing the ISO

18
sensitivity on a camera has one main trade off however, which is the introduction of noise

and grain into the image. Higher quality cameras have less observable ISO noise depending

on the image sensor on the camera and the noise reduction capability of the image

processor.

Preventing Image Blur

It is important to prevent image blur when acquiring images for photogrammetry,

but it can be a difficult task in a low light environment such as an underground mine. A

tripod is a useful tool to prevent camera shake which causes motion blur in photos, however

with a long shutter speed, even the movement of the camera by touching the shutter button

can induce camera shake and cause motion blur. A remote camera shutter or usage of the

10-second programmed timer is helpful when images are being affected by the press of the

shutter. Image stabilized lenses can also help, particularly when hand held or on a monopod

when movement is more likely to occur. Most digital cameras on an automatic setting in a

low light environment will try to balance the exposure triangle by using a high aperture, a

slower shutter speed and increased ISO sensitivity level to have a properly exposed

photograph. If the camera is not on a tripod, the slower shutter speed often results in a

blurred image. Shutter priority mode can be helpful to force the camera to use a faster

shutter speed, however it will requirement the camera to utilize a larger aperture and higher

ISO sensitivity level.

Focal Length

The focal length of a lens for photography refers to the distance, in millimeters,

between the lens and the sensor of the camera. This specification affects the magnification

19
of a lens. A longer focal length, over 50mm on a full-frame camera, will magnify the image.

On the other hand, a smaller focal length will capture more of the scene.

Histograms

A histogram is a graph showing the brightness of a camera’s pixels. A histogram

centered to the left is underexposed, middle is balanced exposure, and on the right is

overexposed. Three example histograms can be seen in Figure 2-11.

Underexposed Balanced Exposure Overexposed

Figure 2-11: Example Histograms Indicating Under-, Balanced and Overexposed Images

Depth of Field

Depth of field refers to the range for the subject to be within focus and can be

increased by stopping down lens aperture as highlighted in Figure 2-12. An increased depth

of field allows more of a subject closer and further from the lens to be in focus. A decreased

depth of field (often referred to as a “shallow depth of field”) reduces the focal plane of the

image. Higher depth of field allows less light into a camera lens while lower depth of fields

allow more light into a lens.

20
 Figure 2-12: Depth of field of a lens (adapted from Wolf, 1974)

Software

Numerous software packages are available to process remote 3D measurements

using photogrammetry. Agisoft Metashape (previously called Agisoft PhotoScan) was the

main software used in this research to compile photogrammetry models. It can be used for

image alignment, mesh reconstruction, decimation for simplification of model mesh sizes,

model texturing, and both point cloud and model exporting in a variety of file formats. The

software supports dedicated graphics card acceleration for image matching, depth map

reconstruction, meshes based from depth maps and texture blending, which is highly

recommended based on the experience of this research. Currently Agisoft Metashape v1.6

21
supports GPU acceleration with Nvidia GeForce GTX 6xx and later graphics cards with

CUDA support as well as AMD Radeon R9 series and later graphics cards with OpenCL

1.1 support (Agisoft, 2019).

RealityCapture was used for its ability to process both laser scans and images

simultaneously into textured 3D meshes. As of the time of writing, the software requires a

Nvidia graphics card with CUDA 2.0+ support and a minimum of 1GB RAM to be able to

create a textured mesh. The recommended specifications prefer at least 4 CPU cores and a

minimum of 16GB or RAM (Capturing Reality, 2020).

Sirovision, a product developed by the Commonwealth Scientific and Industrial

Research Organization (CSIRO – Australia’s national science agency) and Datamine, was

also used, primarily for its native integration with stereophotogrammetry and 3D

geological mapping. The company markets a dual camera system for

stereophotogrammetry designed for underground mine use as well as support for drone-

based photogrammetry of surface mining operations (Datamine, 2020).

3DM Analyst by ADAM Technology is another photogrammetry software

developed in Australia and used by the mining industry. It has also been marketed for

UAV-based photogrammetry in open pit mines (ADAM Technology, 2020). 3DM Analyst

was not available for testing in this research project.

ShapeMetriX 3D is a photogrammetry-based software by 3GSM GmbH in Austria

that is designed for engineering and geology applications without the use of a tripod. The

software has been demonstrated for use in tunneling, surface and underground mining and

geotechnics, for digital face documentation, pillar and rock slope stability assessment,

22
geotechnical data acquisition, deformation monitoring, volumetric calculations and

roughness coefficient estimation (3GSM, 2014). Though the software was not available for

testing in this project, other researchers and consultants have used it for blast design and

analysis in underground limestone mines (McClure, 2019).

Another commonly used photogrammetry software that is designed for drone-based

photogrammetry is Pix4D, which features integration with many popular UAV systems for

autonomous flight and image collection by using GPS navigation, with its own integrated

photogrammetry processing. The software has built in support for calculating and reporting

stockpile volumes, measuring distances and surface areas (Pix4D, 2020). Several surface

crushed stone quarries that were spoken with for this research utilize Pix4D with their UAV

fleets.

A few non-photogrammetry software packages were used in this research,

including CloudCompare, Faro Scene LT, Maptek PointStudio and Adobe Lightroom.

Adobe Lightroom was used to manage and organize image sets collected in this research

for photogrammetry modelling. The software can efficiently handle large numbers of

photos and allows for batch resizing and exporting of groups of photos. Lightroom also

features RAW image support, which was the primary image file type used in this research

(Adobe, 2020). RAW images retain the highest level of quality in digital quality. In

comparison to the JPEG image format, which is a lossy format as it uses compression to

save file space, partially destroying some image detail, particularly in shadows and

underexposed areas of photographs. This extra image detail may be useful to retain in

photos taken in underground mines, but memory storage space should be adequately

allocated to compensate for the increased file sizes (Rouse, 2008).

23
 Faro Scene LT was used to import laser scan data obtained from the Faro 3D

terrestrial laser scanner. The Faro 3D lidar saves 3D point cloud data in *.FLS format,

which is not supported by all 3D point cloud viewing and editing software at the time of

writing. After processing the scans in Scene LT and generating the project point clouds,

the data was converted and exported to an *.E57 point cloud format for use with

CloudCompare, RealityCapture, Metashape and Maptek PointStudio.

CloudCompare is an open source 3D point cloud and triangular mesh processing

software that is capable of importing both point clouds from laser scans as well as

photogrammetry for scaling, alignment and change detection between point clouds

(CloudCompare, 2018). It was also useful for converting point clouds from one file type

to another.

Maptek’s PointStudio software is able to import photogrammetry-based point

clouds as well as laser scan point clouds for a variety of different geotechnical and

engineering tasks. For this research it was utilized for surface change detection between

point clouds such as spalling and deformation. It was also used for aligning multiple laser

scans, creating stereonets from mapped discontinuities, and fragmentation analysis from

point clouds sourced from photogrammetry models of underground limestone mines

captured in this research.

Risk Management

One of the challenges in large opening underground mines is the assessment and

monitoring of ground conditions. Experts in the study of rock mechanics and ground

control have recommended that a “Rock Mass Hazard Risk Index” should be created in

24
addition to a “Roof Fall Risk Map” to help identify areas of potential roof instability

(Esterhuizen, 2008). This is created by periodic visual observations and rating of the roof

conditions in an underground stone mine. The Roof Fall Risk Index (RFRI) combined with

personnel exposure rankings can be used to produce a Roof Fall Risk Map as seen in Figure

2-13. Photogrammetry is a tool that could be used to catalog an enhanced visual

representation of a feature for a database of ground conditions in a mine that can be

analyzed over time for changes.

Figure 2-13: Roof Fall Risk Map

25
Chapter 3

Equipment Design and Methodology

Lighting Considerations

Any underground mine makes for a challenging environment to acquire

photographs, but the large and cavernous openings in an underground limestone mine make

it particularly difficult to capture adequate quality photographs to be used for

photogrammetry. One of the primary considerations for photogrammetry in this type of

environment is lighting.

The images required for photogrammetry need to be evenly lit and exposed

throughout each image. While an onboard camera flash could be used, it is unfavorable

due to the likelihood for the light to reflect off of dust particles in the air and result in an

improperly exposed image (McCartney, 1997). Off-camera flashes help to alleviate this

problem and can provide more lighting power compared to an on-camera flash. However,

considering there is no natural light available for ambient light in underground mines, a

dedicated continuous light was tested to provide the most consistent quality images for the

photogrammetry in this research. Added benefits of a continuous lighting system was its

usefulness to provide ambient lighting to adjust camera settings, set camera focus, and also

provide more overall visibility to reduce slips, trips and falls while moving from one

camera station to the another.

An important feature of the continuous lighting systems tested was the inclusion of

a diffused, or opaque, lens for each light. The broad and evenly cast light helped to

26
distribute light evenly from the center to the corners and edges of each image and reduce

shadows. The miner’s cap lamp was turned off during each image capture as it could also

concentrate a beam of light onto the image subject and negatively influence the camera’s

auto light metering system to underexpose the image on average. Spot beam lights were

tested; however, the beam was too heavily focused on the center of each image, resulting

in overexposure of the center of each photograph and underexposure in the outer sections

as seen in Figure 3-1 below, which also includes histograms for each respective image.

Figure 3-1: (L) Example of Underexposure of the Image due to Spot Lighting versus
(R) Balanced Exposure from Diffused Flood Lighting

Underexposed halos on the outer edges of an image can make photo alignment in

photogrammetry difficult. The software is looking for similar features to match between

photos, so it is important to capture each image fully exposed across the field of view, with

27
the subject sharp and in focus. A lighting system for photogrammetry must be designed to

match with the selected camera and lens for good results in an underground mine.

Lighting Test

Several different lighting sources were tested before finalizing the photogrammetry

rigs used for this project. An LED flashlight with lens diffuser was used in initial

photogrammetry tests underground as seen in Figure 3-2. However, without a mount or an

extra set of hands from a helper, it was difficult to keep the light steady while composing

each photograph and maintain a consistent orientation between photographs.

Figure 3-2: Initial Underground Photogrammetry Testing (image courtesy of Jonathan Baggett)

The other lights tested were a variety of weather-sealed diffused LED’s mounted

to a tripod and powered by fully-charged 11,000 mAh 12V DC Lithium Ion battery packs.
28
Two pairs of off-road LED lights were tested, one pair rated at 30-watts each and the other

50-watts each. The 30-watt LED lights can be seen below in Figure 3-3. A pair of Lume

Cube brand LED’s (version 1) that were used for drone-based photogrammetry in this

research study were also included in the test. In addition, a large 100-watt off-road LED

light bar was tested.

Figure 3-3: 30-watt LED’s with Battery Mounted to 24” Steel Bar

A pair of lumen meters were used to determine the light intensity at varying

distances, from 5 to 20 meters. One of the lumen meters can be seen below in Figure 3-5.

The intensity value recorded was an average between each instrument. The results for each

light tested can be seen in Figure 3-5.

Figure 3-4: Lumen Meter Used for Measuring Light Intensity

29
 Figure 3-5: Brightness of Lights Tested at Various Distances

The results of the lighting test were as expected. The 100-watt LED light bar was

the brightest, followed by the 50-watt LED’s, the 25-watt LED’s, the Lume Cubes, and the

flashlight. However, in practice the 100-watt LED light bar was too cumbersome to carry

and mount along with the camera did not feature a diffuser. The 50-watt LED’s provided

a consistent lighting pattern and were able to be mounted onto a 24-inch fabricated steel

bar capable of being attached to a tripod or monopod. The lighting test also supported the

Inverse Square Law Formula, that the intensity of light is inversely proportional to the

square of the distance:

1
𝐼𝐼 ∝
𝑑𝑑2

Where I = light intensity (candela, W/m2)

d = distance from light source (m)

30
Lens Field of View Test

The camera field of view (FOV) is an important consideration for photogrammetry,

particularly underground. There is limited space between the photogrammetry subject and

camera position. Underground limestone mines can have large entry heights. In some

observed cases, the heights of the mine entries and pillars were over 30-meters tall. In the

case of modelling the full perimeter of a pillar, it is helpful to have high vertical coverage

in each image to reduce the number of passes required in a series of images. It was desirable

in this research to be able to acquire the full height of a ~10-meter tall pillar in circular path

around the perimeter of the pillar. It was also important to ensure that each image would

have adequate exposure by having full lighting across the field of view.

For the test, a range focal length lenses were tested on a full frame DSLR camera

in an underground limestone mine on a single pillar with a height of approximately 10

meters and space to position the camera far enough away to test both wide angle and zoom

lenses from a fixed position as seen below in Figure 3-6.

Figure 3-6: Pillar Photographed for Lens and ISO Tests

31
 The camera was tested with a variety of available lenses in order to determine which

combination provided an appropriate field of view to capture the full 10-meter height of

the pillar, and study the impact of field of view on each image in a real-world underground

mine setting. A Canon 6D DSLR was used for this test, in combination with a Sigma 35mm

f1.4 prime lens, a Sigma 50mm f1.4 prime lens and a Canon 85mm f1.8 prime lens. In

addition, a Canon 16-35mm wide-angle zoom lens and Canon 24-105mm kit lens were

tested. The field of view from a 16mm, 35mm, 50mm and 85mm focal length on a full

frame Canon DSLR are illustrated in Figure 3-7 below. A 35mm prime lens was chosen as

the best fit to pair with a Canon full-frame camera to capture the full height of the 10-meter

pillar in one pass, while still having space between pillars and entries to capture subsequent

images.

Figure 3-7: Determining the Optimal Lens Focal Length

32
 Full-frame cameras are recommended over crop-sensor cameras for

photogrammetry in underground limestone mines, not only for the improved image quality

and low light capabilities from the larger sensor, but mainly because of the effect of the

crop factor. On Canon cameras, the crop sensor has a crop of 1.6x, meaning the effect of a

crop sensor camera paired with a 50mm lens is approximately the same as an 80mm field

of view on a full-frame Canon DSLR. Similarly, a 35mm lens on a crop sensor Canon

DSLR has a similar field of view as a ~56mm lens on a Canon full-frame DSLR. In order

to get the same field of view as a 35mm on a full-frame Canon DSLR, a 22mm lens would

need to be used on the crop-sensor camera. Figure 3-8 on the following page illustrates the

crop factor and how it affects the field of view of an image with a 50mm lens on a full-

frame camera versus a 50mm lens on a crop-sensor camera.

Figure 3-8: Impact of Full-Frame vs Crop Sensor Camera

33
Camera Setting Test

It is important to reduce camera shake in photography, particularly in low light

environments such as underground mines. A modern digital camera has a processor that

can meter the photographic scene and automatically adjust the f-stop, shutter speed, and

ISO setting in an attempt to have a balanced image exposure to the given light available.

On automatic mode, most cameras will generally try to maintain a faster shutter speed to

reduce handheld camera shake at the expense of a combination of high ISO noise and a

low f-stop. Most digital cameras now have special settings for known scenes, such as night

time mode, that is intended for use on a tripod to allow the camera to use a slower shutter

speed to better expose the image. As explained previously, it is important to keep as much

of the photogrammetry subject in sharp focus as possible, which is accomplished with a

higher f-stop over f/5.6 to f/8.0. Due to the low amount of light available underground in a

mine, it is important to learn the manual and assisted setting adjustments in the camera

used for underground photogrammetry.

In order to determine the optimum camera settings for the selected camera and lens

combination in the real-world underground mine environment, a test was conducted to

determine the combination of f-stop and ISO setting required to keep the shutter speed low

enough to reduce camera shake on a monopod. A Canon 6D full-frame DSLR camera

mounted on a tripod with a pair of 50-watt LED lights was used in aperture priority mode

with a Sigma 35mm f1.4 prime lens. The 10-second camera timer was used to prevent

camera shake on the tripod during the test. The lens was tested on the camera with varying

f-stops from f/1.4 up to f/8.0. Camera sensor sensitivity settings were varied in increments

from ISO 100 up to ISO 25,600 for two different f-stops, f/8.0 and f/5.6. Each image was

34
analyzed for noise and the exposure adjustments made by the camera processor were

inspected. A summary chart is below in Figure 3-9 and a table listing the full results can

be seen in Appendix A.

Figure 3-9: ISO vs Exposure at Varying Aperture Settings

Lens Correction

After determining that a 35mm fixed lens provided the best field of view for most

photogrammetry subjects in this research, grid analysis was used to pre-calibrate a lens

correction profile in Agisoft Metashape. Using a tripod to steady the camera, a series of

photos were taken of a 2D checkerboard lens grid displayed on a computer monitor as seen

below in Figure 3-10.

Figure 3-10: Lens Correction by Grid Analysis

35
 The procedure allowed Agisoft to generate an adjusted lens profile that was applied

to correct for slight radial distortion (k1, k2, k3), minor affinity and skew (b1), and the lens

optical axis interception with the sensor plane (cx, cy) as per the Agisoft guidebook

(Agisoft, 2019). The adjusted lens profile for the Canon 6D DLSR in combination with the

Sigma 35mm f1.4 lens can be seen below in Table 1. Graphs showing the distortion versus

radius can be seen in Figure 3-11, indicating greater lens distortion on the outer edges.

Table 1: Lens Correction Profile Created for 35mm Lens

cx: -11.4916

f: 2680.54512 cy: -0.749608

k1: -0.0839391 p1: 0

k2: 0.1528 p2: 0

k3: -0.10491 b1: -0.320276

Figure 3-11: Radial & Decentering Distortion (35mm lens)

36
Monocular Photogrammetry

Equipment Development

After determining the 35mm focal length was optimum for the surveys on a full-

frame camera, an apparatus was developed to make consistent quality images using a single

DSLR camera. It was important to minimize the amount of time required to get set up for

photogrammetric surveys and move from station to station for each image. Another 24-

inch steel bar was drilled and tapped to affix to a standard 1/4”-20 camera / tripod mount.

A camera tripod quick release mount was attached at the top center of the bar, such that the

camera could be attached quickly and securely and with limited available light. 60-watt

diffused LED lights were mounted on opposing ends of the bar and angled to cast light

across the full field of view of the lens. The lights featured weather-sealed male 2-pin

Deutsch connectors, so a wiring harness was created to adapt female 2-pin Deutsch

connectors to each light and connect in parallel to an 18-gauge 2.1mm x 5.5mm DC power

cable that could easily be connected to a portable lithium ion power bank. A summary of

the equipment used can be seen in Table 2.

Table 2: Monocular Photogrammetry Equipment List

Description: Quantity:
Camera: Canon 6D (full frame) 20.2-megapixel DSLR 1
Lens: Sigma 35mm f1.4 DG HSM Art 1
Filter: B+W 67mm XS-Pro Clear Protection Filter 1
Support: Generic Monopod 1
Bracket: 24” steel bar drilled to ¼”-20 1
Lighting: 60-watt 3” diffused beam LED 2
Battery: 11,000 mAh 12V DC Lithium Ion 1
Battery Attachment: 3D Printed Mount Attached to Bracket 1
Camera Quick Release Mount with ¼”-20 standard 1
Attachment: thread

37
 While a tripod produced the steadiest platform to take photographs, it was often

difficult to setup and level quickly on the irregular and often muddy surfaces of the mine

floor, therefore a monopod was used to steady the camera as seen in Figure 3-12. The

subject pillar was developed in a rectangular shape and was mostly flat, allowing for a

shallower depth of field. It was determined to take the images of Pillar A at a f-stop of 5.6

and ISO setting above 6,000 to keep the shutter fast enough to prevent motion image blur.

Figure 3-12: DSLR Photogrammetry Using a Monopod (image courtesy of Roger Lataille)

38
Stereophotogrammetry

Equipment Development

A dual camera photogrammetry rig was constructed such that stereo pairs of images

could be captured simultaneously. The equipment consisted of (2) Nikon D700 FX format

(full frame) DSLR cameras each paired with identical Nikon AF FX 50mm f1.8D

aspherical prime lenses. Tiffen 52mm ultraviolet lens filters were used to protect the lenses

from dust in the mine. The cameras were mounted on opposing ends of a 24-inch steel bar

that was drilled and tapped at ¼-inch x 20 to attach to a Manfrotto 3126 aluminum tripod.

Tripod quick release mounts were utilized for each camera so as to minimize setup in the

mine. A protective watertight case was also modified to accept both cameras, lenses and

accessories to prevent damage when transporting the equipment to the test locations in the

mine as seen in Figure 3-13.

Figure 3-13: Stereoscopic Camera Storage Protection

39
 Cameras were synchronized by affixing SMDV RFN-4s wireless remote shutter

release antennas to the MC30 connection on each camera as seen in Figure 3-14. By setting

both transponders and remotes to the same radio frequency, one remote control could be

used to trigger both cameras simultaneously. The remote shutter release was also important

for eliminating the camera shake that is often induced when manually pressing a camera

shutter. Camera synchronization was checked by taking several stereopair photos of a

smartphone stopwatch and each pair was within five hundredth of a second between left

and right paired photos.

Figure 3-14: Wireless Remote Shutter Control for Stereopair Photogrammetry

(left image courtesy of Michelle Crotto, right image courtesy of SMDV)

Green laser sights were mounted to both cameras to check for camera alignment.

Each laser was attached to the hot shoe mount of each camera by designing and 3D printing

a picatinny rail to hot shoe adapter. Green lasers were chosen over red lasers for improved

visibility. A green laser sight with a wavelength of 532 nanometers is closer to the center

of the visible light spectrum and visible over 90 meters. On the other hand, the wavelength

40
for a red laser sight is higher at 635 nanometers and less visible beyond 20 meters (Optics

Planet, 2016). The laser sights, while not required, did prove helpful for stereo camera

alignment, particularly in the dark mine environment. The lasers were also useful for

aligning subsequent stereo pairs to the 3x3 camera gridlines for overlap between camera

stations. A photograph of the laser sight and hot shoe adapter can be seen in Figure 3-15.

Figure 3-15: Laser Sight with Hot Shoe Adapter to Assist Camera Alignment

Lighting was achieved by also mounting two 3-inch diffused 30-watt flood beam

light emitting diode (LED) lights to the steel bracket holding the cameras. The lights, which

were designed for off-road automotive use, were selected for this application as they were

affordable (US$60/pair), ruggedized to withstand underground mine use, IP68-rated

waterproof, passively cooled and featured built-in diffusers for a soft and broad flood light

pattern. The lights each featured six 5-watt LED’s with each light having a current draw of

2.0 amps at 12-volts via a sealed 2-pin male Deutsch plug connector. The lights were rated

to produce 2600 lumens each with a daylight color temperature of approximately 5000

41
degrees Kelvin. An image of the stereoscopic photogrammetry rig can be seen in Figure

3-16.

Figure 3-16: Stereophotogrammetry Rig

The lights were powered by a rechargeable 12-volt lithium ion battery rated at

11,000 mAh and a maximum output of 6 amps. The battery, which had its own plastic

protective housing and toggle power switch, weighed 498 grams. A mounting case was

designed and 3D printed to securely attach the battery to the stereo camera bracket. An 18-

gauge 2.1mm x 5.5mm DC power adapter cable was used for a quick connection to the

battery, and spliced, soldered and shrink wrapped to a Y-adapter cable with each end

soldered and crimped to sealed 2-pin female Deutsch plug connectors to connect directly

42
to each LED light. A summary of the equipment used to test stereo pair photogrammetry

follows in Table 3.

Table 3: Stereoscopic Photogrammetry Equipment List

Description Quantity:
Camera: Nikon D700 (full frame) 12.1-megapixel DSLR 2
Lens: Nikon 50mm f1.8 2
Filter: Tiffen 52mm UV Protection Filter 2
Support: Manfrotto 3126 Aluminum Tripod 1
Bracket: 24” steel bar drilled to ¼-20 1
Lighting: 30-watt 3” diffused beam LED 2
Remote Trigger: SMDV RFN-4s wireless remote 2
Battery: 11,000 mAh 12V DC Lithium Ion 1
Battery Attachment: 3D Printed Holster Strapped to Bracket 1
Camera Attachment: Quick Release Mount 2
Alignment: Green Dot Laser Sight 2
Misc: 3D Printed Picatinny Rail to Hot Shoe Adapter 2

Image Naming Convention for Stereo Pairs

The TIFF image type was used to preserve image detail, particularly in low light

areas which can be lost with image compression filetypes such as JPEG. Camera file

settings were modified in-camera such that individual photographs were named

corresponding to their camera position. By default, Nikon DSLR photograph files are

named “_DSC####”, with the # symbols reserved for camera numbering. Using the Nikon

file naming setting within the shooting menu, the three-character name for each image was

modified such that images taken from the left camera in the stereo pair were prefixed with

“_LEF####.tif” while images taken from the right camera were prefixed with

“_RIG####.tif”. The photograph number count was reset for both cameras such that each

camera was on the same image count, i.e. “_LEF0001.tif” and _RIG0001.tif”. The cameras

43
were checked periodically to ensure both cameras remained in sequence. After

downloading the images from both cameras to a computer and project folder, a batch file

was written and used to restructure photograph filenames in the folder with the number as

the prefix and left and right indication as the suffix. This allowed the image sets to be

ordered in left and right pairs (0005_LEFT.tif, 0005_RIGHT.tif, 0006_LEFT.tif,

0006_RIGHT.tif, 0007_LEFT.tif and so on). This procedure simplified the process of

matching stereo image pairs when on the computer in preparation for processing in the

Sirovision photogrammetry software.

Computational Processing Time

An important consideration for implementing photogrammetry in a mine’s

workflow is the computer on which the processing is done. While nearly any modern

desktop computer or laptop can process a photogrammetry model, a higher end personal

computer that can utilize a dedicated graphics card for processing may be a worthwhile

investment if results are desired more quickly. Throughout the course of this research, all

models were compiled on two primary computers:

1. A personal computer with a 3rd generation Intel i7 processor, 32 GB RAM

and dedicated AMD Radeon R9 280 graphics card with 3 GB of VRAM,

built in early 2013, and running Windows 10 64-bit.

2. A personal computer with a 9th generation Intel i9 processor, 32 GB RAM

and dedicated Nvidia GeForce RTX 2080 Ti graphics card with 16 GB of

VRAM, purchased in late 2018, and running Windows 10 64-bit.

44
 For this test, a set of 300 images of the building seen in Figure 3-17 was collected

and processed using the same settings within Agisoft Metashape on a variety of desktop

Figure 3-17: Photogrammetry Model of a Building

and laptop computers. The image set was collected using a DJI Mavic 2 Pro UAV flown

manually around 3 vertical sides of the inner courtyard of the building as seen in Figure

3-17. The drone features a Hasselblad L1D-20c camera with a pixel size of 6.44 x 6.44 μm

that was used to capture the images. The images were recorded at a focal length of

10.26mm (28mm in terms of 35mm equivalent), F-stop of F/5, ISO 100, and shutter speed

of 1/200. Image and depth map thumbnails can be seen in Appendix C.

Figure 3-18: Location of Images Used for Photogrammetry Model

45
 Using Agisoft Metashape (version 1.6.2), the images were used to create the

photogrammetry model from Figure 3-17. A batch file was created to process the point

cloud, mesh and texture map in sequence without interruption. The photos were aligned

using medium accuracy settings, a low-quality dense cloud, a triangular mesh sourced from

the dense cloud using low quality settings and medium face count to improve to speed at

which the model could be compiled for the test. Higher quality settings would greatly

increase the time in which the model would be processed. The point cloud contained

218,246 points with a coverage area of 0.247 km2, resolution of 61.9 cm/pix, and point

density of 2.61 points/m2. The RMS reprojection error was 0.13596 pix, with a maximum

reprojection error of 0.4134 pix and an average tie point multiplicity of 6.025. The newer

desktop computer with a 9th generation Intel i9 processor and Nvidia 2080 Ti graphics card

was able to complete all processing steps in 5 minutes and 9 seconds. The older desktop

computer with a 3rd generation Intel i7 processor and AMD R9 280 graphics card was able

to complete all processing steps in 30 minutes and 26 seconds. This represents a major

productivity improvement for which the photogrammetry processing can occur, with a time

savings of 25 minutes and 17 seconds, or in 17% of the time as the older computer. The

older computer took nearly six times as long.

The same 300 images were also processed by students on 20 other computers, both

desktops and laptops, using identical batch processing settings in Agisoft Metashape. The

processing speed varied widely between the different computers as seen in Figure 3-19.

46
 Figure 3-19: Benchmark Processing Time for Building Model

The fastest machines utilized Intel 9th generation processors, but more importantly,

featured Nvidia 2080 graphics cards that the software could use for accelerated processing

for image matching, depth map reconstruction, meshes and texture blending.

It is therefore very important to consider the computer on which the

photogrammetry models will be compiled. A relatively small investment in computing and

graphics processing power can make great returns in the speed and efficiency of the model

computation.

47
Photogrammetry Workflow

The most important step when creating a photogrammetry model is the image

collection. Using the respective camera and LED light rigs for both monocular and stereo-

photogrammetry, a similar image collection pattern was practiced. Agisoft recommends

capturing images in sequence with a 60% overlap between adjacent photos (Agisoft, 2020).

An image collection pattern for a typical pillar in an underground limestone mine is

illustrated below in Figure 3-20. A flowchart for creating a photogrammetry model in

Agisoft can be seen in Figure 3-21 on the following page.

Figure 3-20: Photogrammetric Survey Around a Typical Mine Pillar

48
 Capture Images

Image Alignment &

Generate Sparse Cloud

Generate Dense Cloud

Create Mesh

Create Solid

Create Texture Map

Figure 3-21: Photogrammetry Workflow

49
 Once the images are captured and downloaded to a computer, they can then be

imported into the Agisoft Metashape software. Images should not be cropped before

importing to Agisoft, and after inspection, unnecessary images should be removed. Agisoft

can also import video and extract image frames from the video to use for photogrammetry.

After the images are imported into the software, they need to be aligned. The software will

search for common features between images as was illustrated in Figure 2-9 and created a

sparse cloud. From the generated sparse cloud, the model region can be adjusted to the area

of interest. A dense point cloud should then be compiled based on the area of the sparse

cloud that was defined. Using the dense cloud as a reference, Agisoft can build a polygonal

mesh based upon the target face count that the user defines. After the mesh is created, a

texture map blended from the images that formed the model can be applied to the surface

(Agisoft, 2019).

If structural detail is needed in a model, it is recommended that a high face count is

used to create the mesh. A high face count does increase the amount of computational time

required and also the amount of processing required to display the model. If the model just

needs to be visualized, a lower face count can be used, or the previous mesh can be

decimated to a lower target face count. With many photogrammetry models, there is no

significant loss of visual information from a decimated mesh, yet the amount of processing

required is greatly reduced as the computer does not have to render as many three-

dimensional polygons that represent the mesh (Agisoft, 2019).

50
Chapter 4

Applications

Mine Risk Databases

As the underground coal mining industry is much larger than underground

limestone, it has had considerably more research and journal publication on risk

assessment. Similar to underground stone mines, roof and rib falls are also the basic ground

control hazards in underground coal. Dr. Alexander Garcia presented an approach to

estimate ground control risk with visual assessments cataloged in Microsoft Excel at the

31st International Conference on Ground Control in Mining that could also be very useful

to management at underground limestone mining operations. An excel spreadsheet can be

created to support the continuous assessment mine roofs with objective risk ratings along

defined intervals, which helps to identify support problems early and allowing for better

management of ground control risks (Garcia, 2012). Photogrammetry can help by

providing a record of ground conditions at a moment in time that can be added to a database

of underground observations with use of data visualization techniques and compared at a

later date for deformation and degradation. Software such as Maptek’s PointStudio

program allows for saving multiple models as historical snapshots of as-built mine

infrastructure for archiving (Maptek, 2020).

51
Mine Headings

As shown in Figure 4-1, this approach can also be used to make a permanent record

the conditions of mine faces as they were active and to monitor their condition if left over

time, using point cloud change detection.

Figure 4-1: Mine Heading Photogrammetry Model Screenshots, Headings 1 through 4

52
Change Detection

By comparing surfaces generated from properly referenced point clouds over time,

changes due to spalling and/or deformation can be detected. In Figure 4-2 below, images

were collected for compiling into photogrammetry models before and after hand scaling of

a heading. The photogrammetry point clouds were produced using Agisoft Metashape on

the same settings for both models. Each model was exported as an *E57 point cloud from

Metashape and imported into CloudCompare for alignment. After alignment, the point

clouds were exported again for use in Maptek PointStudio for change detection. The point

clouds were downsampled to improve the processing speed for surface triangulation. Using

the change detection tool in PointStudio, the surfaces were compared to one another. The

analysis tool was able to calculate the depth differences between points clouds in the

locations where hand scaling occurred, represented in blue in the Figure below.

Figure 4-2: Change Detection between Photogrammetry Models After Hand Scaling

53
 Mine headings can also be analyzed for blast efficiency. By having a three-

dimensional point cloud representation of the rock face, the surface can be inspected for

over-breakage and under-breakage after blasting. Photogrammetry also allows for a

quantification of the volume of material created by the blast. Software such as Maptek’s

PointStudio program has routines to determine over and under-breakage, calculate

volumes, as well as a tool to provide fragmentation analysis by inspecting the size

distribution of broken rock from a blast that is useful for making blast improvements and

optimizations (Maptek, 2020). Using the point clouds generated by photogrammetry, it is

possible to use these point cloud computational analysis tools to inspect blast fragmentation

observed in Figure 4-3 below.

Figure 4-3: Photogrammetry Model Analyzed for Blast Fragmentation

54
Karst Mapping

The karst voids commonly encountered in underground limestone mines present

unique challenges for the miners and mining engineers to design ground control. Irregular

surfaces are difficult to measure by traditional surveying methods. Photogrammetry can be

helpful for mapping and designing ground control as millions of three-dimensional surface

points can be obtained that can also be used to make volumetric calculations and references

for designing custom supports to prevent fall of ground. Figure 4-4 below is a photo of a

photogrammetry survey of a karst in an underground limestone mine.

Figure 4-4: Photogrammetry Survey of an Open Karst Void

In this application, 69 photos were taken from multiple viewpoints of a karst

feature. The photogrammetry model, compiled in low quality settings in Agisoft, was based

on those images with a pixel size of 6.58 x 6.58 µm. 78,085 tie points were generated with

216,735 projections and a reprojection error of 0.936 pix. The mesh was calculated with a

55
face count of 154,588 before the texture was applied as seen in the side, frontal and upward

views of the photogrammetry model featured in Figure 4-5.

Figure 4-5: Karst Model – (CW) Side, Frontal and Upward Views

Ground Control

A challenge for mine mining engineers is designing and monitoring ground control

to remediate dangerous ground conditions. Photogrammetry allows for the operator to

collect geospatial information of the mine workings from a safe distance and recreate them

in a digital model for further analysis and design. Underground limestone operations use

various forms of roof control such as rock bolting, mesh, shotcrete, and steel sets to

strengthen and shield the roof and rib in an effort to prevent ground falls. Photogrammetry

can be used to determine the amount of shotcrete installed in any given visible location

whereby before / after photogrammetry models are created. Similarly, rock bolt or steel set

56
movement could be detected by having accurate photogrammetry-based point clouds of the

engineered object analyzed over time. An example of a drone-based photogrammetry

survey of a mine pillar supported with rock bolts and synthetic mesh can be seen below in

Figure 4-6. A screenshot of the 3D point cloud representation of a section of the rock bolts

and mesh can be seen in Figure 4-7.

Figure 4-6: Drone Photogrammetry Survey of a Pillar with Rock Bolt and Mesh Reinforcement

Figure 4-7: Drone Photogrammetry Model of Rock Bolt and Mesh Reinforced Pillar

57
As-Built Underground Mine Construction Surveys

Not only is photogrammetry helpful for mapping out the irregular surface features

of an underground mines, but it is also useful for documenting mine construction projects

with as-built surveys. An engineer may utilize photogrammetry to map out the dimensions

and volume of a concrete form in an underground structure, similar to the example in Figure

4-8 below. After the installation is complete, the as-built model can be used for monitoring

the installation and also as a virtual finished blueprint example that mine workers can

reference for future projects.

Figure 4-8: Photograph of an Underground Construction Project

58
 As-built photogrammetry surveys can be especially important for monitoring large

ground control installations underground. Figure 4-8Figure 4-6 is a photogrammetry model

produced from 126 photos of a steel set ground control installation under a large grouted

karst opening. The model features 132,208 tie points with 349,684 projections and a

reprojection error of 0.561 pix. The model was generated with a high-quality mesh with a

face count of 12,296,500 polygons.

Figure 4-9: As-Built Photogrammetry Survey of Steel Set Ground Support

Pillar Monitoring

The mine pillars in an underground room and pillar mines should be inspected for

spalling and deformation over time. Similar to the change detection application for mine

headings, photogrammetry and laser scanning of mine pillars are useful for generating

three-dimensional digital point cloud representations of a mine features that can be

geospatially compared for changed over time in software such as PointStudio.

Photogrammetric models of several mine pillars were generated from a variety of

DSLR camera and lens focal length combinations, in addition to photogrammetry models

produced from images collected by UAVs underground. Figure 4-10 and Figure 4-11 are

59
screenshots of the three-dimensional photogrammetry models obtained from a Canon 6D

DSLR + 35mm f1.4 prime lens and a DJI Mavic 2 Pro UAV, respectively, of the full

perimeter of the same pillar measuring approximately 10 meters tall in an underground

limestone mine.

Figure 4-10: Fully Modelled Mine Pillar Using Full Frame DSLR + 35mm Lens

Figure 4-11: Fully Modelled Mine Pillar via Drone-based Photogrammetry

The DSLR-based photogrammetry model was composed of 127 photos with a pixel

size of 6.58 x 6.58 µm. The model was constructed with 67,712 tie points, 333,813

projections with a reprojection error of 0.328 pix and a face count of 54,786. The drone-

based photogrammetry model, on the other hand, had a face count of 929,763 polygons.

The pixel size was 2.41 x 2.41 µm with an image count of 89 photos. The model featured

60,351 tie points with 154,172 projections and a reprojection error of 1.17 pix.

60
 These models were created to test loop closure and to compare versus laser

scanning. While no change was detected between tests using the same camera, these

models could serve as base references in the future for pillar deformation or change studies.

Geotechnical Analysis

A benefit of successful utilization of photogrammetry in underground limestone

mines is having a three-dimensional visualization tool that also has high spatial accuracy.

Commercial software is available such as Datamine’s Sirovision software that is designed

for the mining industry to help map discontinuities and create stereo plots, rose plots, and

contour plots from photogrammetric imagery recorded in mines (Datamine, 2020). In this

research, photogrammetry point clouds from Agisoft Metashape were exported to a *.OBJ

file and then imported to Maptek’s PointStudio software. Figure 4-12 is a sample of

mapped discontinuities and a stereonet generated from a drone-based photogrammetry

model of a pillar. Remote sensing such as laser scanning and photogrammetry can allow

for virtual mapping of geological features without having to be physically near a potentially

dangerous ground control hazard.

Figure 4-12: Geotechnical Mapping of a Point Cloud Obtained from Photogrammetry

61
Highwall Inspections

An important part of any underground mine is the mine entrance. Subjected to

frequent traffic, mine personnel pass by and through a mine portal on a daily basis, and

may or may not take the proper amount of time to inspect the conditions of the highwall

above. Additional exposure to the elements and fracturing by the influence of freeze and

thaw can also increase the likelihood of a rock fall at the mine entrance.

The mine portal of one of the underground limestone mines visited was modeled

with photogrammetry based on images collected with a UAV as seen in Figure 4-13. Using

a commercially available DJI Mavic 2 Pro with a 35mm equivalent focal length of 28mm,

143 photographs were captured of the highwall with a pixel size of 2.41 x 2.41 µm. The

photogrammetry model featured 93,309 tie points and 510,444 projections with a

reprojection error of 3.87 pix. The mesh face count was decimated to 100,000 polygons.

Figure 4-13: Mine Entrance Highwall Modelled via Drone-based Photogrammetry

62
Smartphone Photogrammetry

To illustrate the accessibility of photogrammetry to a modern mining operation, a

smartphone was also tested to collect survey images. It must be noted that in this particular

test, a large diesel-powered utility light was being used in the vicinity, illuminating the

subject beyond that of a normal area of an underground mine. Nonetheless, using an Apple

iPhone 6, which features a 35mm equivalent focal length of 29mm and pixel size of 1.52

x1.52 µm, 57 images were collected of a pillar face as well as the roof and floor in front of

the pillar. The model compiled 62,323 tie points and 214,500 projections, with a face count

of 979,895. The total reprojection error was 0.877 pix.

Figure 4-14: Underground Mine Photogrammetry from a Smartphone

63
Chapter 5

A Comparison of Laser Scanning and Photogrammetry

in an Underground Limestone Mine

Richard Bishop, Juan Monsalve, Jon Baggett, Aman Soni, Nino Ripepi

Adapted from a conference proceeding published and presented at the 2019 Society of
Mining, Metallurgy & Exploration Annual Conference (SME 2019) in Denver, Colorado.
Used with Permission.

(ABSTRACT)

Technology plays an ever-increasing role in improving the safety and efficiency of

mining operations. Laser scanning and photogrammetry are two useful methods for

capturing 3D digital representations of real-world objects. While both technologies have

been applied to the mining industry in numerous ways, the practical applications in an

operating underground limestone mine has been tested for this paper, including for

visualization and site characterization. Each technology is capable of creating highly

detailed geospatial point clouds, but are all point clouds created equal? This paper presents

a comparison of a laser scan and photogrammetry of a limestone pillar and addresses the

strengths and limitations of each method for creating digital models of operating

underground limestone mines.

64
Introduction

The United States crushed stone business is a ~$14 billion per year industry

representing 1,430 companies which operate 3,782 mining operations across the 50 states.

In 2017 alone, 1.33 billion tons of crushed stone was produced with 76% used

predominately for road construction and repairs, but also for the production of cement,

lime, chemical and agricultural uses. Over 70% of domestic crushed stone producers source

their material from limestone and dolomite deposits, with currently only 82 mines (2%)

operating underground (USGS, 2018).

Underground mining requires a different technical skillset, and poses great

challenges for production that must be both safe and economical. Therefore, there are many

risks that an operation must mitigate in order to produce underground limestone. Some of

the challenges in underground limestone mining involve geological hazards that are unique

to the rest of the industry. The geologic features such as faults and weaknesses can be

difficult to assess in a production environment (NIOSH, 1998). The consequences of

failure due to improperly designed underground mining ground support systems, such as

from a roof failure, are not only life threatening for the miners themselves, but can easily

shut down an entire mining operation.

Underground limestone mines in the eastern U.S. have become more common over

the past decade and typically there is less underground experience resulting in a need for

more engineering controls. Over the past ten years 40% of underground mining fatalities

were caused by ground control issues related to ground collapses. Over the same period

the underground stone mining industry has had the highest fatality rate in four of those ten

years, more than any other mining sector (MSHA, 2016).

65
 Photogrammetry and laser scanning are technologies that can help characterize rock

mass discontinuities and joint sets, as well as help monitor ground control. They have also

proven useful for 3D mapping. These technologies have the potential to help mine

operators better characterize ground failure mechanisms and visualize hazards.

Overview – Laser Scanning

Three-dimensional laser scanning, or LiDAR, is a remote imaging method which

can collect 3D point coordinates of a scanned scene with high resolution and accuracy. It

also helps with detailed mapping of the structural features present in underground

workings, not only by increasing the safety and precision over traditional surveying with

automated scans, but also by reducing time while mapping (Adu-Acheampong, et al.,

2013). It is well-suited for underground use as it requires no lighting to complete a scan.

Software such as Maptek’s I-Site program are designed to process the point clouds obtained

from laser scanning, which can be used for geotechnical analysis in underground limestone

mining operations (Monsalve et al, 2018). In these applications, laser scanning has also

been shown to be effective in determining the volumetric changes when measuring rib

displacement, and particularly suited for determining sloughage off of a surface (Slaker,

Westman, & Fahrman, 2013). The 3D laser scan can be seen in Figure 5-1.

66
 Figure 5-1: 3D point cloud of the face of subject limestone pillar from laser scanning

Overview - Photogrammetry

Photogrammetry is a method of using photographs taken from different locations

to derive measurable spatial relationships within the subject. Most mining operations in the

USA presently have access to a digital camera in addition to a computer capable of the

basic processing requirements to compile a photogrammetry model. However,

photogrammetry it is not without unique challenges in an underground mine environment.

Photography requires adequate light to record photographic data. While the low light

environment encountered in an underground mine can be offset with a camera flash and/or

supplementary lighting as well as longer exposure times, dust, moisture and the large

cavernous openings in underground limestone mines can make it difficult to get quality

surveys. Figure 5-2 illustrates the use of photogrammetry to model the perimeter of the

subject pillar, the face of which will be used in comparison to the laser scan presented in

Figure 5-1.

67
 Figure 5-2: 3D rendering of subject pillar using photogrammetry

Equipment and Methodology

Photogrammetry

While photogrammetry can be performed with virtually any camera, equipment

selection is important for efficient quality surveys in large opening underground mines

such as room & pillar limestone mines. For this comparison test, a Canon 6D full-frame

digital single-lens reflex (DSLR) camera was paired with a Sigma 35mm f1.4 prime lens.

This camera contains a 20.2-megapixel CMOS sensor capable of 5472 x 3648 pixel images

and a wide ISO sensitivity range from 50 – 102,400.

A more important consideration than the camera body, is the choice of camera lens.

The focal length of a camera lens refers to the optical distance between the lens and camera

image sensor and is the specification that describes if a lens provides magnification (zoom)

or a wide-angle view. Figure 5-3 illustrates the effect of focal length on field of view for a

full-frame camera with various focal length lenses.

68
 Figure 5-3: Effect of focal length on field of view

On a full-frame camera, 50mm is regarded as the focal length equivalent to what

our eyes see, whereas 35mm is considered the initial range of wide-angle lenses and 80mm

would be considered a zoom lens. A crop-sensor camera paired with a 35mm lens is similar

to the field of view of a full-frame camera with a 50mm lens and a 50mm lens on a crop-

sensor camera has a similar field of view as an 80mm lens on a full-frame camera

(Vorenkamp, 2015).

Photogrammetry requires overlapping images in order for the processing software

to calculate a point cloud. 60% overlap is the general guideline recommended between

images (Agisoft, 2019). In order to save time and simplify the collection of photographs,

the height of the subject pillar should fill the camera frame, such that the survey can be

performed laterally along the face of the pillar.

69
 The subject pillar in our test measured approximately 13 meters in height during

the first phase of mine development when the survey was performed. With a 13-meter entry

between the pillar face and the rib, the distance that away from the subject to make

photographs is a limiting factor. In this survey, a 35mm lens allowed for the full height of

the pillar to fit vertically in the camera frame while also allowing for adequate lighting

mounted alongside the camera tripod to produce a consistently illuminated subject within

the field of view of an individual photograph as seen in Figure 5-4.

Lighting is a very important consideration for photogrammetry surveys in

underground limestone mines. After much trial and experimentation, a pair of 3-inch by 3-

inch diffused flood beam 20-watt LED’s were chosen and powered by a 11,000 milliamp-

hour 12-volt lithium ion rechargeable battery pack. This configuration allowed for

consistent lighting during the survey and clear images with the camera set to a f-stop of

f/6.7 with a 2 second exposure at ISO 1000.

Figure 5-4: Photogrammetry survey of mine pillar

70
Laser Scanning

The laser scan was performed with a Faro Focus3D laser scanner. According to the

manufacturer, the instrument is capable of 360o scanning from 0.6 meters up to 120 meters

with a step size of 0.009o. The unit is capable of capturing up to 976,000 points per second

with a vertical field of view of 305o (Faro, 2013).

The laser scanner was mounted to a tripod during the survey in two locations

located 13 meters apart corresponding with the approximate entry width in front of the

pillar. These locations are indicated by the two circular gaps in the point cloud floor in

Figure 5-1: 3D point cloud of the face of subject limestone pillar from laser scanning and

by the tripod icons in the map in Figure 5-5.

Figure 5-5: Survey overview map (icons not to scale)

71
 The laser scanning process was mostly automated. Reference objects were placed

between the two scan locations. The two separate scans were then merged into a single

point cloud by the alignment of the reference objects. The scan settings were specified on

the built-in menu at ¼ resolution and started with the press of a touchscreen icon. In less

than 6 minutes per station, the device automatically rotated in all directions collecting

geospatial point locations from the reflection of the laser onto the surrounding surfaces.

Figure 5-6 shows the laser scanner on the tripod in preparation for the first scan of the mine

pillar.

Figure 5-6: Laser scanning of a limestone mine pillar

Data Processing & Analysis

Post-processing is typically the most time intensive portion of a photogrammetry

survey, particularly on higher quality settings. The photogrammetry survey processing was

72
completed in Agisoft PhotoScan Standard (version 1.4.4) on a Windows 10 laptop with an

Intel i7-5500U processor @ 2.40 GHz, 16 GB RAM and GeForce 840M graphics. Camera

matching and alignment took 2 minutes 50 seconds and 44 seconds, respectively. The dense

point cloud consisted of 44,865,485 points and was reconstructed using ultra high-quality

settings and aggressive depth filtering which required 3 hours 21 minutes to generate the

depth maps and 2 hours 28 minutes to generate the dense cloud. At this stage the model

was exported as an *.E57 point cloud file for comparison with the laser scan data.

The laser scan data was processed and aligned with the Faro Scene software.

Processing required 2 minutes 34 seconds on the same computer. The laser scans acquired

21,646,505 points between the two full 360o surveys on the ¼ resolution setting. It was

also exported as an *.E57 point cloud.

Both photogrammetry and laser scan point clouds were then imported into

CloudCompare (version 2.10-alpha) for analysis. CloudCompare is a free, open-source, 3D

point cloud processing program that allows for comparison of two dense point clouds

(CloudCompare, 2018). After importing, the laser scan point cloud was already scaled and

registered, however the photogrammetry point cloud required resizing and alignment /

registration with the laser scan. The laser scan point cloud was then segmented to reduce

the area scanned to the pillar face for comparison. The laser scan and photogrammetry

point clouds can be seen individually in Figure 5-7.

73
 Figure 5-7: Point Clouds – Laser Scan (top) vs Photogrammetry (bottom) from CloudCompare

The quality of point clouds from both laser scanning and photogrammetry are more

than sufficient to map discontinuities in geotechnical software packages such as Maptek I-

Site. The additional color information provided by the photogrammetry model can be

helpful to better visually identify mineral variation and also provide a more photorealistic

rendering of the mine area for visualization. Figure 5-8 is a distance comparison from

CloudCompare between the laser scan point cloud and the photogrammetry point cloud on

a point by point basis, including a frequency distribution.

74
 Figure 5-8: Distance error analysis (using CloudCompare)

In the proceeding images, CloudCompare revealed that the greater deviation

between scans occurred at the corners of the pillar corresponding to the edges of the

surveys. Both surveys had more dense point clouds in the center of the pillar where there

were more overlapping data points. The photogrammetry survey had more overlapping

images at the center and the two laser scan stations converged point clouds at the middle

of the pillar.

75
 The results support the quality of both scans and the ability to merge both point

clouds when required. For example, if a mining operation does not own their own laser

scanner, but may have had a consultant scan part of the mine, they could generate a new

point cloud using photogrammetry to append the laser scan with additional information. In

addition, texture detail from the photogrammetry model can be overlaid on the laser scan

point cloud to provide more photorealistic detail of the working face or mine pillars to

better visualize a short term mine plan and/or ground control design.

Advantages & Limitations

Laser Scanning

Laser scanning has several advantages over photogrammetry in an underground

limestone mine. A laser scanner is able to acquire a detailed point cloud of the full

surrounding area around the station, seamlessly merging the sides, roof and floor. This was

found to be a more difficult task with photogrammetry as it requires overlapping images

that become more difficult to orientate when capturing overhead. The laser scanner also

performs very well in a dark environment, whereas photogrammetry needs adequate

lighting to ensure proper camera focus and sharp images. Also, point clouds from the laser

scanner are already scaled when they are generated, whereas photogrammetry needs

reference objects for scaling after processing. It also has the advantage of being quicker to

process after a large survey.

Photogrammetry

Photogrammetry has the advantage of having a lower upfront cost for equipment,

much of which might already be owned by the mining operation. If a camera, lens, battery

or light needed to be replaced, there are many available options on the market. The software
76
for creating photogrammetry models is constantly advancing, as is computing power.

However, one of the limitations of photogrammetry over laser scanning is processing time

relative to laser scanning for complex scenes. Small surveys can be performed and

processed quickly, but larger, more expansive surveys allow more room for user error

during the survey and more processing time. Good lighting is essential for good results in

photogrammetry. Thus, some areas such as the upper portion of limestone pillars and the

roof, can be over 30 meters overhead and be challenging to illuminate. Also, some areas

may have limited access, such as under unsupported ground and open karsts, which may

impact the completeness of a survey.

Merging Both Technologies

Both photogrammetry and laser scanning have their unique advantages and

disadvantages, however opportunity exists to enhance the quality of each survey by

combining them. Figure 5-9 shows the face of the pillar (on the right-hand side) merged

onto the 360o laser scan using RealityCapture software (Capturing Reality, 2020). The

detailed color texture map from photogrammetry provides added detail and the laser scan

provides orientation. This approach is very useful for visualization and can be used for

training exercises in a photorealistic virtual reality (VR) environment.

77
Figure 5-9: Merged laser scan (L) + photogrammetry (R)

78
Chapter 6

Conclusions and Future Work

Conclusions

Underground limestone mining is a growing segment of the U.S. mining industry,

which is associated with inherent roof fall hazards that must be controlled and managed

during the life of mining. The industry faces many risks due to geological hazards with

major consequences of failure, including worker injuries, production delays / setbacks,

miner safety, lost reserves, flooding and unanticipated roof control costs, amongst many

others.

The methodologies and techniques for monitoring underground limestone mine

workings, such as photogrammetry and laser scanning, show great promise as equipment,

computational power, and software continues to advance. The survey detail both

technologies can provide can help operations be proactive with risk management and

ground control design and monitoring. It was found in this study that both point clouds

compare favorably in the level of detail that can be used for mapping geological structures

and discontinuities as well as a base for ground control design using the scans as as-built

surveys.

With limestone ore being a relatively low value commodity compared to coal and

precious metals mining, cost and budget considerations when trying to implement new

technologies can be a major obstacle for companies. However, using risk assessment

concepts to quantify the consequences of failure of reacting to ground failures after they

79
occur rather than using the best available monitoring and modelling techniques, a

compelling argument can be made for maintaining a level of acceptable service in

underground limestone mine ground control by preventing roof and pillar failures.

Photogrammetry can be applied in numerous ways to improve the safety and

visualization of large opening underground limestone mines, including permanent

documentation of the rock face, as-built surveys, change detection, modelling and

visualization of karst features, blast analysis, in addition to many other applications. The

challenging lighting conditions typical of underground mine environments can be

overcome with recent advancements in LED lighting technology and availability of

inexpensive lightweight lithium batteries.

Applied research has indicated that the selective pairing of quality, yet affordable,

DSLR cameras and high-speed fixed lenses with focal lengths capable of capturing

balanced exposures of the photogrammetry subject from overlapping perspectives can

yield accurate three-dimensional point clouds. In addition, the proper camera and lens

combination together with a well thought out lighting system can be used to capture

millions of data points in one photo that can be combined with subsequent overlapping

images to produce digital models of the underground mine environment. Camera stations

can be quickly set up ensuring coverage of large areas with low cost equipment that is easy

to obtain and within reach of the average mine operation.

It was also observed that areas with limited access can impact the quality of a

photogrammetry survey. It is important to try and prevent shadows in photographs as they

can interfere with the model calculations including alignment, but also leave holes within

the mesh that take away from the accuracy of the model. In addition, the importance of

80
good lighting cannot be overstated. Full diffused illumination of the camera field of view

produces images that are much easier to align, preventing underexposure around the edges

of each photograph and the elimination of overexposed sections of an image due to spot

lighting. Aerial drones with onboard lighting can also be used successfully underground to

complement the photogrammetry workflow and gain exposure to areas that may otherwise

be too dangerous to access.

Another consideration for implementing photogrammetry into the workflow at

working mine operation is the computer chosen for processing. A small investment in a

modern personal computer with a dedicated graphics card can lead to substantial time

savings when compiling photogrammetry models. While it was demonstrated that a

smartphone can be used for photogrammetry, it is recommended to make an investment for

a DSLR camera with exposure settings that can be adjusted. Understanding of the manual

override settings of a digital camera can greatly improve the reliability of photogrammetry

surveys by reducing image blur, controlling excessive ISO noise, as well as preventing out

of focus images.

As underground mines embrace new remote sensing technologies to enhance the

collection of measurable survey data, software tools are expanding in their capability to

process large geospatial data sets. These tools can be utilized to improve the safety of

underground limestone operations by adding to the ability to increase quantifying metrics

of the effectiveness and condition of ground control. In addition, these digital observation

and measurement tools can be useful to optimize the mine development process.

Photogrammetry has been demonstrated as one such remote sensing method that can be

applied in underground stone mines. As more operations adopt three-dimensional mine

81
mapping and mine planning, photogrammetry can provide an additional level of detail

beyond traditional underground surveying methods to help operations better visualize and

assess ground fall risks and ground control design.

Future Work: Virtual & Augmented Reality Applications

Virtual reality is a computer-generated simulation that creates an immersive

environment, whereas augmented reality superimposes layers of computer-generated

sensory information onto the real world. With the recent introduction of low cost and

portable head-mounted displays for virtual reality such as the Oculus Quest and many

smartphones featuring augmented reality applications, virtual reality and augmented reality

technologies are becoming more accessible and gaining in popularity.

This research work has demonstrated the ability of photogrammetry to enhance

underground mine mapping and help with the visualization of underground mine

development and hazards that may be difficult to see with the naked eye and limited

lighting. Outside of mining, one of the common applications of photogrammetry is the

creation of computer-generated imagery (CGI), including photorealistic environments and

characters for video games. The models created in this research can be used as a virtual

mine environment for virtual reality. A photorealistic virtual mine environment could help

supplement mine safety and health training with interactive learning modules, mine hazard

recognition training, enhanced visualization of as-built mine surveys, in addition to many

other applications. Early examples of photogrammetry models from this research being

used for testing in virtual reality and augmented reality can be seen below in Figure 6-1

and Figure 6-2, respectively.

82
 Figure 6-1: Virtual Reality for 3D Visualization of Underground Mine Models

Figure 6-2: Augmented Reality of a Mine Pillar

Future Work: Autonomous Drones

Many underground mines have been in operations for decades. Some areas of mine

development may not have been adequately mapped, which is a risk for many mines due

to potential flooding and unknown ground conditions. Access to these areas may be limited

and dangerous for traditional surveying techniques.

83
 The primary challenges with using drones in underground mines relates to

positioning, drift and collision avoidance. Most commercial and consumer UAV’s are

designed for flights on surface where Global Positioning Satellite (GPS) signals are

available for positioning and collision avoidance sensors rely upon ambient light. GPS

signals do not penetrate well into underground mines and optical collision avoidance

sensors fail due to lack of light, therefore UAV’s must either be flown manually, or

programmed with Simultaneous Localization and Mapping (SLAM) algorithms utilizing

sensors such as lidar, to map and position the drone within an unknown environment. A

benefit of lidar-based SLAM on a drone is the ability of the technique to both help the

UAV avoid collisions and also create a map that can be used for other purposes.

Initial testing of photogrammetry with drones in underground limestone mines has

been promising. The challenging lighting conditions for drone-based photogrammetry have

been addressed in this research with high-powered LED’s. Figure 6-3 on the following

page illustrates an application in an underground limestone mine where a high-roof

underground section would be difficult to collect photographs for photogrammetry from

the ground. A drone has an advantage over ground-based photogrammetry as it is able to

ascend parallel to a mine pillar and maintain consistent lighting and camera depth of field

at a fixed distance perpendicular to the face from the ground to the mine roof.

Future work is recommended to implement SLAM and specialized collision

avoidance sensors tailored for the underground mine environment onboard UAV’s. A

drone programmed to fly autonomously through an unknown mine area to map and collect

data would be able to prevent the additional risk to mine personnel from entering a

potentially hazardous unknown environment.

84
Figure 6-3: Drone with LED’s Mapping a 100-foot Tall Entry with Photogrammetry

85
Bibliography
3GSM. (2014). ShapeMetriX 3D - Measurement and assessment of rock and terrain
surfaces by metric 3D images. Graz, Austria.
ADAM Technology. (2020). 3DM Analyst. Retrieved from http://www.adamtech.com.au/
Adobe. (2020). Adobe Lightroom. Retrieved from
https://www.adobe.com/products/photoshop-lightroom.html
Adu-Acheampong, A., Byers, T., Dodd, M., Basson, F., Bartlett, M., & Duran, J. (2013).
An Overview of Imaging Technologies at Newmont Mine Sites. Denver, CO:
SME.
Agisoft. (2019, October 8). Metashape Software. Retrieved from Agisoft LLC.:
http://www.agisoft.com/
Agisoft. (2020, May 17). Image Capture Tips - Equipment and Shooting Scenarios.
Retrieved from Support - Tips and Tricks: https://www.agisoft.com/support/tips-
tricks/
Baggett, J. (2019). A Study of Ground Penetrating Radar Methods in an Underground
Stone Mine to Improve Ground Control [Masters thesis, Virginia Tech].
Blacksburg, VA: Institutional Repository at Virginia Tech.
Bishop, R. E., Monsalve, J., Baggett, J., Soni, A., & Ripepi, N. (2019). A Comparison of
Laser Scanning and Photogrammetry in an Underground Limestone Mine. SME
Annual Conference. Denver, Colorado: SME.
Capturing Reality. (2020). RealityCapture Software. Retrieved from Capturing Reality
S.R.O.: www.capturingreality.com
CloudCompare. (2018, October 31). Retrieved from CloudCompare Open Source Project:
https://www.danielgm.net/cc/
Datamine. (2020). Sirovision - Open Pit/Underground Mapping & Analysis. Retrieved
from https://www.dataminesoftware.com/solutions/sirovision-geotechnical-
mapping/
Faro. (2013). Faro Focus3D - Features, Benefits & Technical Specifications. Retrieved
from Faro: https://www.faro.com/
Haycocks, S. (1993). Surface Versus Underground Mining: A Risk Analysis Approach.
In Risk Assessment/Management (pp. 3-7). Blacksburg, VA: Virginia Tech.
Kehew, A. E. (1988). General Geology for Engineers. Englewood Cliffs, NJ: Prentice
Hall.
Lachambre, S., Lagarde, S., & Jover, C. (2017, 06 23). Photogrammetry Workflow.
Retrieved from Unity: https://unity3d.com/files/solutions/photogrammetry/Unity-
Photogrammetry-Workflow_2017-07_v2.pdf
Lanmar Services. (2014). Laser Scanning vs. Photogrammetry. Retrieved from
http://lanmarservices.com/2014/11/07/laser-scanning-vs-photogrammetry/

86
Maptek. (2020, May 8). PointStudio Software. Retrieved from MAPTEK Pty Ltd.:
https://www.maptek.com/products/pointstudio/
McCartney, S. (1997). Mastering Flash Photography: a course in basic to advanced
lighting techniques. New York: Amphoto Books.
McClure, R. (2019). Underground 3-D Photogrammetry, Blast Design, Analysis and
Rockmass Characterization. 2019 Underground Stone Safety Seminar. Louisville,
KY.
Monsalve, J. J., Baggett, J., Bishop, R., & Ripepi, N. (2019). Application of Laser
Scanning for Rock Mass Characterization and Discrete Fracture Network
Generation in an Underground Limestone Mine. International Journal of Mining
Science and Technology, 131-137.
MSHA. (2016). Summary of Fatal Accidents at Metal/Nonmetal Mines with Preventative
Recommendations. Washington D.C.: U.S. Department of Labor.
NIOSH. (1998). Innovative Hazard Recognition Training for Underground Limestone
Miners. U.S. Department of Health and Human Services, National Institute for
Occupational Safety and Health.
NIOSH. (2018). Centers for Disease Control and Prevention: NIOSH Mining. Retrieved
from NIOSH Mine and Mine Worker Charts: https://wwwn.cdc.gov/NIOSH-
Mining/MMWC/MineMap/
Optics Planet. (2016, September 13). Retrieved from Optics Planet:
https://www.opticsplanet.com/howto/how-to-choose-a-laser-sight-red-or-green-
laser-sight.html
Pix4D. (2020). Drone mapping and 3D modeling for mining. Retrieved from
https://www.pix4d.com/
Rouse, A. (2008). Understanding RAW Photography. Photographers' Institute Press.
Slaker, B. A., Westman, E., & Fahrman, B. (2013). Determination of Volumetric
Changes from Laser Scanning at an Underground Limestone Mine. SME Annual
Meeting. Denver, CO: SME.
USGS. (2012). Limestone - A Crucial and Versatile Industrial Mineral Commodity. U.S.
Department of the Interior, U.S. Geological Survey.
USGS. (2018). Mineral Commodity Summaries. Washington D.C.: U.S. Department of
the Interior, U.S. Geological Survey.
USGS. (2020). Mineral Commodity Summaries. Washington D.C.: U.S. Department of
the Interior, U.S. Geological Survey.
Vorenkamp, T. (2015). Understanding Crop Factor. Retrieved from BH Photo Video:
https://www.bhphotovideo.com/explora/photography/tips-and-
solutions/understanding-crop-factor
Wolf, P. R. (1974). Elements of Photogrammetry. New York: McGraw-Hill.

87
 Appendix A: Photogrammetry Model Summaries

Subject Camera Focal Photos Tie Points Pixel Size Projections Reprojection Face Count
Length Error

(mm) (pix)
(#) (#) (µm) (#) (#)

Pillar-A Canon 6D 16mm 28 13,518 6.58 x 6.58 35,165 0.348 343,712

Pillar-A Canon 6D 35mm 127 67,712 6.58 x 6.58 333,813 0.328 54,786

Pillar-A DJI Mavic 2 Pro 28mm 89 60,351 2.41 x 2.41 154,172 1.17 929,763

Pillar-A Face Canon 6D 35mm 34 17,233 6.66 x 6.66 111,350 0.552 149,999

Pillar-A Face Nikon 700 Stereo 50mm 48 33,086 8.46 x 8.46 171,820 0.511 490,216

Pillar-A Face DJI Mavic 2 Pro 28mm 35 43,790 8.14 x 8.14 140,744 0.251 451,133

Pillar-B DJI Mavic 2 Pro 28mm 119 99,760 2.41 x 2.41 244,537 0.790 6,359,879

Building DJI Mavic 2 Pro 28mm 300 174,888 2.41 x 2.41 1,031,908 0.788 105,227

Highwall DJI Mavic 2 Pro 28mm 143 93,309 2.41 x 2.41 510,444 3.87 100,000

88
 Subject Camera Focal Photos Tie Points Pixel Size Projections Reprojection Face Count
Length Error

(mm) (pix)
(#) (#) (µm) (#) (#)

Karst Canon 6D 35mm 69 78,085 6.58 x 6.58 216,735 0.936 154,588

Steelsets Canon 6D 35mm 126 132,208 9.99 x 9.99 349,684 0.561 12,296,500

Heading 1 Canon 6D 35mm 44 48,594 6.58 x 6.58 157,951 1.19 603,382

Heading 2 Canon 6D 35mm 79 76,193 6.58 x 6.58 276,015 1.18 755,853

Heading 3 Canon 6D 35mm 42 46,225 6.58 x 6.58 149,740 1.17 533,704

Heading 4 Canon 6D 35mm 47 59,515 6.58 x 6.58 176,382 0.929 653,413

Mine Rib iPhone iPhone 6 29mm 57 62,323 1.52 x 1.52 214,500 0.877 979,895

89
Appendix B: Camera Setting Test Data

Table 4: Exposure Time at Various Aperture and ISO Settings

F-Stop Exposure ISO Image F-Stop Exposure ISO Image

(sec) # (sec) #
1.4 0.17 800 1047 5.6 0.17 16,000 1039
2.0 0.30 800 1046 5.6 0.13 20,000 1040
2.8 0.70 800 1045 5.6 0.10 25,600 1041
3.5 1.00 800 1044 6.7 30.00 100 1002
4.0 1.50 800 1043 8.0 20.00 200 1003
5.6 20.00 100 1021 8.0 10.00 400 1004
5.6 10.00 200 1022 8.0 8.00 640 1005
5.6 6.00 400 1023 8.0 6.00 800 1006
5.6 6.00 500 1024 8.0 6.00 1,000 1007
5.6 4.00 640 1025 8.0 3.00 1,600 1008
5.6 3.00 800 1026 8.0 3.00 2,000 1009
5.6 3.00 1,000 1027 8.0 2.00 2,500 1010
5.6 2.00 1,250 1028 8.0 1.50 3,200 1011
5.6 1.50 1,600 1029 8.0 1.50 4,000 1012
5.6 1.50 2,000 1030 8.0 1.00 5,000 1013
5.6 1.00 2,500 1031 8.0 0.70 6,400 1014
5.6 0.70 3,200 1032 8.0 0.70 8,000 1015
5.6 0.70 4,000 1033 8.0 0.50 10,000 1016
5.6 0.50 5,000 1034 8.0 0.30 12,800 1017
5.6 0.30 6,400 1035 8.0 0.30 16,000 1018
5.6 0.30 8,000 1036 8.0 0.25 20,000 1019
5.6 0.25 10,000 1037 8.0 0.17 25,600 1020
5.6 0.17 12,800 1038

90
Appendix C: Photogrammetry Image Sets

Pillar A (Full-frame DSLR 35mm)

Thumbnails 1 of 4:

91
Pillar A (Full-frame DSLR 35mm) - Thumbnails 2 of 4:

92
Pillar A (Full-frame DSLR 35mm) - Thumbnails 3 of 4:

93
Pillar A (Full-frame DSLR 35mm) - Thumbnails 4 of 4:

94
Pillar A (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 4:

95
Pillar A (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 4:

96
Pillar A (Full-frame DSLR 35mm) – Depth Map Thumbnails 3 of 4:

97
Pillar A (Full-frame DSLR 35mm) – Depth Map Thumbnails 4 of 4:

98
Pillar A (Full-frame DSLR 35mm)

Thumbnails 1 of 1:

99
Pillar A (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 1:

100
Pillar B (UAV)

Thumbnails 1 of 4:

101
Pillar B (UAV) - Thumbnails 2 of 4:

102
Pillar B (UAV) - Thumbnails 3 of 4:

103
Pillar B (UAV) - Thumbnails 4 of 4:

104
Pillar B (UAV) – Depth Map Thumbnails 1 of 4:

105
Pillar B (UAV) – Depth Map Thumbnails 2 of 4:

106
Pillar B (UAV) – Depth Map Thumbnails 3 of 4:

107
Pillar B (UAV) – Depth Map Thumbnails 4 of 4:

108
Karst (Full-frame DSLR 35mm)

Thumbnails 1 of 2

109
Karst (Full-frame DSLR 35mm) – Thumbnails 2 of 2

110
Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 2

111
Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 2

112
Steel Sets (Full-frame DSLR 35mm)

Thumbnails 1 of 3

113
Steel Sets (Full-frame DSLR 35mm) – Thumbnails 2 of 3

114
Steel Sets (Full-frame DSLR 35mm) – Thumbnails 3 of 3

115
Steel Sets (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 3

116
Steel Sets (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 3

117
Steel Sets (Full-frame DSLR 35mm) – Depth Map Thumbnails 3 of 3

118
Pillar Face (UAV)

Thumbnails 1 of 1

119
Pillar Face (UAV) – Depth Map Thumbnails 1 of 1

120
Building (UAV)

Thumbnails 1 of 11

121
Building (UAV) – Thumbnails 2 of 11

122
Building (UAV) – Thumbnails 3 of 11

123
Building (UAV) – Thumbnails 4 of 11

124
Building (UAV) – Thumbnails 5 of 11

125
Building (UAV) – Thumbnails 6 of 11

126
Building (UAV) – Thumbnails 7 of 11

127
Building (UAV) – Thumbnails 8 of 11

128
Building (UAV) – Thumbnails 9 of 11

129
Building (UAV) – Thumbnails 10 of 11

130
Building (UAV) – Thumbnails 11 of 11

131
Building (UAV) – Depth Map Thumbnails 1 of 11

132
Building (UAV) – Depth Map Thumbnails 2 of 11

133
Building (UAV) – Depth Map Thumbnails 3 of 11

134
Building (UAV) – Depth Map Thumbnails 4 of 11

135
Building (UAV) – Depth Map Thumbnails 5 of 11

136
Building (UAV) – Depth Map Thumbnails 6 of 11

137
Building (UAV) – Depth Map Thumbnails 7 of 11

138
Building (UAV) – Depth Map Thumbnails 8 of 11

139
Building (UAV) – Depth Map Thumbnails 9 of 11

140
Building (UAV) – Depth Map Thumbnails 10 of 11

141
Building (UAV) – Depth Map Thumbnails 11 of 11

142
Heading 1 (Full-frame DSLR 35mm)

Thumbnails 1 of 2

143
Heading 1 (Full-frame DSLR 35mm) – Thumbnails 2 of 2

144
Heading 1 (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 2

145
Heading 1 (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 2

146
Heading 2 (Full-frame DSLR 35mm)

Heading 2 (Full-frame DSLR 35mm) – Thumbnails 1 of 3

147
Heading 2 (Full-frame DSLR 35mm) – Thumbnails 2 of 3

148
Heading 2 (Full-frame DSLR 35mm) – Thumbnails 3 of 3

149
Heading 2 (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 3

150
Heading 2 (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 3

151
Heading 1 (Full-frame DSLR 35mm) – Depth Map Thumbnails 3 of 3

152
Heading 3 (Full-frame DSLR 35mm)

Thumbnails 1 of 2

153
Heading 3 (Full-frame DSLR 35mm) – Thumbnails 2 of 2

154
Heading 3 (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 2

155
Heading 3 (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 2

156
Heading 4 (Full-frame DSLR 35mm)

Thumbnails 1 of 2

157
Heading 4 (Full-frame DSLR 35mm) – Thumbnails 2 of 2

158
Heading 4 (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 2

159
Heading 4 (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 2

160
Mine Drift with Karst (Full-frame DSLR 35mm)

Thumbnails 1 of 9

161
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 2 of 9

162
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 3 of 9

163
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 4 of 9

164
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 5 of 9

165
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 6 of 9

166
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 7 of 9

167
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 8 of 9

168
Mine Drift with Karst (Full-frame DSLR 35mm) – Thumbnails 9 of 9

169
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 1 of 9

170
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 2 of 9

171
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 3 of 9

172
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 4 of 9

173
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 5 of 9

174
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 6 of 9

175
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 7 of 9

176
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 8 of 9

177
Mine Drift with Karst (Full-frame DSLR 35mm) – Depth Map Thumbnails 9 of 9

178
Mine Rib (iPhone 6)

Thumbnails 1 of 2

179
Mine Rib (iPhone 6) – Thumbnails 2 of 2

180
Mine Rib (iPhone 6) – Depth Map Thumbnails 1 of 2

181
Mine Rib (iPhone 6) – Depth Map Thumbnails 2 of 2

182
Appendix D: Permission for Release

SME 2019 Preprint (19-095)

183
184