Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Proceedings of the ASME 2018 13th International

Manufacturing Science and Engineering Conference


MSEC2018
June 18-22, 2018, College Station, TX, USA

MSEC2018-6453

A Laser Triangulation-based 3D Measurement System


for Inner Surface of Deep Holes
Zhu Ye, Wang Lianpo
Department of Precision Machinery and Precision Instrumentation
University of Science and Technology of China
Hefei, Anhui, China
Yonggang Gu, Zhai Chao
Experiment Center of Engineering and Material Science
University of Science and Technology of China
Hefei, Anhui, China
Baoyang Jiang,Jun Ni
Department of Mechanical Engineering
University of Michigan
Ann Arbor, Michigan, USA

ABSTRACT is required to be non-contact, high-efficiency, 3D-capable and


Basing on the laser-triangulation principle, a deep hole have a small diameter.
inner surface inspection system is developed in this research. The necessity and difficulty of inner surface measurement
Compared with existing inner surface measurement systems, together make it a hotspot in the field of optical inspection.
this research takes assembly errors and refraction distortion into Researchers have been working on this topic for years, and
consideration and proposes a flexible laser plane calibration several practical measurement systems based on different
technique based on binocular vision that is easy to operate. The technologies have been developed. Albertazzi [1] proposed an
measurement system calibrated by proposed method performs inner surface measurement system using fringe projection
well in the field test in which the maximum absolute error of technology. A halogen lamp and a transparent tube is used
the measured ring gauge diameter is 3μm. Experiment indicates where a set of sinusoidal helical lines are printed to generate
that this system is accurate and suitable for inner surface sinusoidal fringe, and the tube is driven by a stepper motor to
measurement. Point cloud registration may be applied in the conduct phase shifting. The structure of this system is
future to extend the system’s applicability. complicated and systematic error originates from various parts,
including the manufacturing error of the sinusoidal helical line
and motion error of the stepper motor, and such error is hard to
INTRODUCTION eliminate. Masayuki [2,3] developed a digital holograph-based
Parts with deep holes, such as gun barrels, air cylinders, inner surface measurement system that can reach sub-micron
and pipelines, are widely used in industry. The working resolution and accuracy. However, digital holograph is a
condition of these parts is directly determined by the quality relative measurement technology, which means that it can only
and precision of their inner surface, especially when shaft-hole obtain ‘height difference’ instead of real 3D shape. In other
match is involved. Inner surface defects can happen in words, the digital holograph-based system can’t measure the
machining process (unqualified machining) or long-term diameter of a cylinder. Inari [4] proposed a laser-triangulation-
running (abrasion), therefore inner surface measurement is based system that uses a conical mirror to generate circle-
necessary in both production test and system maintenance. structured laser beam. This kind of measurement system has
However it is hard to measure the inner surface because of its been widely studied by other researchers these years because of
characteristics: closed shape, narrow space and full contact with its great advantages, such as simple and compact construction,
the shafts. Thus, a practical inner surface measurement system

1 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


high measurement speed, and full-field 3D reconstruction with an additional camera, to observe a blank planar board at a
ability. few (at least two) different unknown orientations. The
additional camera is pre-calibrated under the detector’s
A typical structure of the laser triangulation-based inner surface coordinate system. Experiments indicate that the mathematical
detector consists of a laser diode, a conical reflector, a CCD model and calibration method are effective, and the system
camera with wide-angle lens, and a stepper motor that drives could reach a high precision with these improvements.
the scanning motion. The system works as follows: The laser This paper is organized as follows: In Section 2, the principles
beam is reflected by a conical reflector and transformed into of measurement and system calibration are introduced in detail,
radial beam , also known as circle-structured light. The radial and system errors caused by assembly error and influence of
beam intersects with the inner surface to create a close-curving refraction in the glass tube are analyzed. In Section 3,
light stripe that can be captured by the CCD camera. The calibration and measurement experiments are presented to
captured images are then transmitted to a computer. Basing on prove the reliability and accuracy of our system and method. In
the principles of close-range digital photogrammetry, the Section 4 the source of remaining errors and further
coordinates of illuminated inner surface points are calculated improvements are discussed. Finally, Section 5 presents the
using the images and the parameters of the CCD camera and conclusion of this work.
radial beam. This detector is driven by a servo motor to scan
the entire inner surface and to reconstruct a complete 3D map, METHODOLOGIES
in which the axial coordinate is obtained from servo feedback. 1. Principles of inner surface measurement
To compensate the assembly error and the machining error of Fig.1 shows the schematic diagram of our measurement
the conical reflector, reliable calibration is required to obtain system that consists of a laser diode, a conical reflector, a CCD
the parameters of the radial beam. camera and a glass tube that connects them together.
Basing on this construction and principle, researchers have
developed several types of practical inner surface detectors.
Zhuang [5,6] performed theoretical analysis on the detector
performance and improved the system resolution to 100μm.
She calibrated the system by measuring a standard ring gauge
that requires precisely alignment with the detector’s axis, and
this turned out to be a difficult operation. Zhang [7] developed
a system in which the laser and the conical reflector are
separated and stretched into the pipe from opposite directions.
Such improvement is beneficial to reduce system error due to
the fact that the radial beam is refracted once by the glass tube,
instead of twice as other systems. He calibrated the system by
using a thin thread to intersect the radial beam at different
locations to create several light spots, and the coordinates are Fig.1 Structure diagram of laser triangulation-based detector
measured by two electric theodolites. Then these points are As illustrated in Fig.2, the camera coordinate system is
used to calculate a conic surface that is regarded as the radial taken as the local coordinate system O-XYZ, the origin of the
beam. This method can reach a very high accuracy (about 30μ coordinates is set at the camera optic center O, and the Z axis is
m) but its operation is very complicated. Wakayama [8-10] along the optic axis.
improved the system by using fibers to transmit laser and
images, and thus the diameter of detector was reduced to 5mm.
The detector with a smaller diameter got wider application in
small-diameter pipe inspection. The proposed mathematical
model did not take assembly errors and machining errors into
consideration, and the parameters are obtained through accurate
assembly and calculation instead of calibration.
Basing on the radial beam technology, an inner surface
measurement system was developed with improved
mathematical modeling and corresponding flexible calibration
technique. The mathematical model takes multiple types of
system errors into consideration, including assembly errors of
the conical reflector and influence of refraction in the glass
tube. In the mathematical model, the beam is regarded as a
plane with four degree-of-freedom under the camera’s Fig.2 System coordinate system and pinhole model
coordinate system. The calibration technique is based on
improved binocular vision and only requires the detector, along

2 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig.2 also illustrates the pin-hole model that describes the and its image projection, e.g. Formula(2), requires to be revised
relationship between the 3D object point P(X,Y,Z) and its to:
image projection p(u, v): 𝑟′ cos 𝜃 𝑢−𝐶𝑥 𝑟′ sin 𝜃 𝑣−𝐶𝑦
= , =
X u  Cx , Y v  C y (1) { 𝑍 𝐹 𝑍 𝐹
(4)
  𝑑𝑔𝑙𝑎𝑠𝑠 sin(𝜔+𝛿)
Z F Z F 𝑟 − 𝑟 ′ = cos 𝛿 cos 𝜔
Where F is the focal length, and ( C x , C y ) is the principal
The symbols are explained in Fig.3.
point. The outgoing beam is regarded as an ideal plane with
For simplicity of calculation, O-XYZ is replaced by a four degree-of-freedom in the camera coordinate system.
cylindrical coordinate system. Taking r as the polar radius Formula (5) is the expression of the outgoing beam, in which
and  as the polar angle, formula(1) is transferred into A B C D are parameters to be calibrated by the
formula (2): proposed algorithm
And thus formula(1) is transferred into formula (2):
r cos  u  C x , r sin  v  C y
 : Ax  By  Cz  D  0 (5)
  (2)
Z F Z F As mentioned above, the point under test P belongs to both
A wide-angle lens is necessary to ensure that the light imaging ray l and structured-light ω . Therefore, the
stripe is completely captured, but this will cause a distortion coordinates of point P can be calculated using generalized
problem. A fourth-order radial distortion model and a second- optical triangulation, which involves solving equation set that
order tangential distortion model as formula (2) are adopted to consists of Eq. (4) , Eq(5) if the corresponding parameters are
describe and correct the distortion. determined.
ud  4 u 
The parameters in Formula (1) are calibrated through
   1  k1r  k 2 r    d
2 Zhang’s camera calibration method [11], and the parameters in
  vd  v  Formula (4) are calibrated by the method proposed in

 d  2k3uv  k 4 r  2u 
 2 2
(3) Section2.2.
As illustrated in Fig.1, the detector is driven by a stepper
 k3 r  2v   2k 4uv 
  2 2  motor and corresponding leadscrew nut mechanism to conduct

F Cx Cy 
and
k1 k2 k3 k4  are together
scanning along the Z axis. The points acquired at different
locations are unified in the global coordinate as a 3D point
called instinct parameters and calibrated through Zhang’s cloud by using coordinates in Z direction, which are offered by
method[11]. feedback of the leadscrew nut. All kinds of geometric
parameters of the inner surface, e.g. machining errors and
All the image coordinates used below are the corrected surface defects could be obtained by analyzing the point cloud.
image points. 2. Principles of system calibration
A blank planar board intersects with the radial beam in an
Beside lens distortion, there ’ s refraction distortion
arbitrary unknown orientation and position, in which the board
introduced by the glass tube as shown in Fig.3.
is regarded as a plane  i . The intersection line is a spatial line
called Li as described by Formula (6)
 i : ai x  bi y  ci z  d i  0
 (6)
  : Ax  By  Cz  D  0
Li is captured by the two cameras, one is the detector
camera illustrated in Fig.1 and the other one is only for
calibration.
According to the principles of projective geometry, a line
corresponds to another line in the central projection. Therefore,
the images that correlate with Li called li1 (in camera1) and
li 2 (in camera2) are also lines, as described by Formula (7), in
which the coordinates under the image coordinate system.
Fig.3. Schematic diagram of the refraction distortion.  li1 : ai1u  bi1v  ci1  0
 (7)
li 2 : ai 2u  bi 2v  ci 2  0
Obviously, the distortion of incident beam changes with
different radius of measured points, while the outgoing beam
could still be regarded as a conical surface before arriving the
inner surface. Therefore, the relationship between object point

3 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Where ai1 bi1 ci1  could be obtained by linear using Zhang’s method [11]. All parameters with the subscript
fitting of extracted central points of the light stripe image. “2” belong to the coordinate system of camera 2. Parameters
of Formula (1) are determined by choosing one image points
Given the focal length F1 of camera 1, the spatial on , thus corresponding object point could be obtained by
expression of li1 in the global coordinate system is generated solving Formula (9) and Formula (10), namely intersection of
as follows: Ai and L pik . Then a set of object points Pik k  1 ~ n
ai1 x  bi1 y  ci1  0 are obtained by choosing several(n) different image points and
 (8)
 z  F1 conducting intersection calculation between Ai and different
A spatial plane is constructed using the optical center (also imaging ray L pik that corresponds to the chosen image points.
the origin point) O and . Formula (9) is the expression of Ai . The planar board is then moved and intersects with the
structured light at several(m) different unknown orientations
ai1F1 x  bi1F1 y  ci1 z  0 (9)
and positions to create different spatial lines then a set of non-
As shown in Fig.4 , Li is the intersecting line of three collinear points P ik k  1 ~ n,i  1 ~ m are obtained by
planes:  ,  i and Ai . Therefore, the points on Li are all the intersection calculation between different plane Ai and its
on the structured-light plane .
Li also correlates with li1 on the image plane of camera
corresponding imaging rays L pik k  1 ~ n.
Then the parameters of laser plane (A B C D) are
2. This correlation implies that each image point pik of li1 calculated by minimizing the square sum of distance defined by
correlates with a point P that belongs to Li and Ai . Formula (11).
Therefore, P can be determined by the intersection of its S d  A B C D   
m m
 Axik  Byik  Cik  D 2 (11)
k 1 i 1 A2  B 2  C 2  D 2
imaging ray L pik with Ai , as shown in Fig. 4.
The circle-structured light calibration is completed.

EXPERIMENTS
1. Construction of the measurement system
The measurement system has been completely built up. As
shown in Fig. 5, it is set up on an optical table and consists of a
detector, which generates the structured light and acquires the
images, and corresponding motion mechanism that actuates the
detector to move along Z axis. The detector consists of a fiber
laser, a conical reflector and a high-precision monochrome
CCD camera with a wide-angle lens. The motion mechanism
consists of a leadscrew nut mechanism with 4mm lead and a
stepper motor whose original resolution is 200 steps per round,
which means that the resolution of the entire drive institution is
0.02mm.

Fig. 4. Schematic diagram of the intersection algorithm


Formula (10) shows the expression of , the imaging ray of
image point p(u2 , v2 ) .
 x2 u2  C x 2 y2 v2  C y 2
  , 
 z2 F2 z2 F2 (10)
x y z   Rx y z T  T
T Fig. 5. Physical map of the measurement system
 2 2 2 The captured images are acquired by a high-speed PCI-E
where (R T) is the coordinate conversion from camera2 to image acquisition card and then processed in LabVIEW, which
camera1 and obtained through binocular camera calibration also controls the stepper motor through a motion control card.

4 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


2. Calibration experiments
As shown in Fig. 6, the calibration experiment system
consists of the measurement system, a blank planar board and
an additional CCD camera.

Fig. 8 .Line extraction and fitting result of the light stripe


Fig. 6. Set-up of the calibration experiment
The calibration procedure is summarized as follows: By choosing huge amount of image points on li 2 ,
(1)Calibrate the two cameras using Zhang’s algorithm [16]. corresponding object point set are obtained through intersection
The external parameters of camera 2 must be transformed into algorithm proposed above. The fitting result is presented in Fig.
the image coordinate system of camera 1. 9 and the obtained parameters are presented in Tab.1.
(2)Block the circle-structured light with the blank planar
board in an arbitrary orientation to create a light stripe, which
will be captured by the two cameras.
(3)Change the orientation and position of the plate and
repeat step (2) five times.
(4)Process the data using the algorithm in the previous
section.
A sample captured image in calibration is presented in Fig.
7.

Fig. 9. Laser plane fitting result


Tab.1 Parameters of the laser beam
Parameters A B C D
Values 0.007 0.006 1 119.87

3. Measurement experiments
Actual measurement experiments must be conducted to
evaluate the accuracy of the calibration procedure. As shown in
Fig. 10, the experimental system is composed of the calibrated
Fig. 7. Image of the light stripe captured in calibration experiment system and 6 ring gauges with different diameters, ranging
As mentioned in the previous section, the expression of from 60mm to 100mm.
li1 is required to determine Ai . Therefore, these light stripes
need to be fitted into lines before the intersection calculation.
Fig.8 shows the extraction and fitting result of Fig7.

5 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 10 Set-up of the measurement experiment
In this experiment, the internal diameters of the gauges are
measured by our system and are compared with the reference Fig.11. Extraction of the light stripe
values. The inner surface of the ring gauges are ideal cylinders. Based on the calibration result in Section 3.2 and the
Therefore, the reference value can be obtained easily through captured images above, the 3D coordinates of the inner surface
3D coordinate measurement machine and cylinder fitting. points can be calculated using the optical triangle method
The detector is driven by the leadscrew nut and the stepper described in formula(6). The measured point cloud of 100mm
motor to carry out linear motion with a step length of 0.02mm, ring gauge is shown in Fig.12.
which means that the CCD camera takes a light stripe image
every 0.02mm. Fig.10 shows one of the captured light stripe
image.

Fig. 12. Point cloud acquired in the measurement experiment


The inner surface points are then fitted to a cylindrical
surface. The diameters of the fitted cylindrical surfaces are
considered the measured values of the ring gauge being tested.
The reference values are provided by a Daisy8106-type three-
coordinate measuring machine with an indication error of 2 μ
m. Each gauge is measured 5 times to evaluate the system’s
repeatability. The measurement results of 6 ring gauges are
presented in Tab.2, where ‘measurement ’ indicates the
Fig. 10. Captured image in the experiment
average measurement value
Then the center of the light stripe is extracted through Tab.2 Measurement results of 6 ring gauges
grayscale function fitting, with the procedures illustrated as
Reference/mm 59.980 69.983 75.981
following:
Measurement/mm 59.982 69.986 75.978
1.Fit the circle using all points with gray levels that exceed
RMS Error/μm 16 25 25
the threshold;
2.Calculate the normal directions on sampling points; Reference/mm 79.979 89.974 99.972
3.Determine cross sections, using the sampling points and Measurement/mm 79.979 89.975 99.975
the corresponding normal directions; RMS Error/μm 14 17 22
4.Fit the gray levels on each cross sections to a smooth As illustrated in the table, the absolute measuring errors
curve, e.g. grayscale function; of diameters are all less than 3μm, and the average RMS error
5.Take the derivative of each grayscale function, and take of all 6 measurements is 19μm.
the extreme points as the extracted center points.
Fig.11 shows a sample of the light stripe center extraction. DISSCUSSION
So far, the system and corresponding calibration method are
proved to be reliable and effective for inner surface diameter

6 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


measurement, while there are still some problems remain to be With these improvements, this system reaches high
solved in the future work. precision and accuracy, and is suitable for on-site measurement.
First, single-point measurement accuracy, which is Some aspects remain to be worked on in the future, such as
significant to defect detection and machining error detection, accuracy evaluation for single-point measurement and point
has not been evaluated due to lack of high-precision reference cloud registration algorithm.
instruments such as theodolites that can measure 3-D
coordinates of a single point. Alternative methods needs to be ACKNOWLEDGMENTS
applied to evaluate the system’s single-point measurement This research was financially supported by National
accuracy and ability to detect defects. For example, we can cut Natural Science Foundation of China [U1331201].
grooves with different known 3D dimension on the sample and
figure out the smallest defect size that the system could detect. REFERENCES
Second, the length of the deep hole is restricted by the [1] Albertazzi Jr A G, Hofmann A C, Fantin A V, et al.
range of the leadscrew,and it is very difficult and expensive to Photogrammetric endoscope for measurement of inner
extend the length. A potential solution is to utilize a segmented cylindrical surfaces using fringe projection [J]. Applied optics,
scanning method based on point cloud. The brief procedure is 2008, 47(21): 3868-3876.
as following: 1. Scan a section of the inner surface and obtain a [2] Yokota, Masayuki, and Toru Adachi. "Digital
3D point cloud; 2. Move the detector to a deeper unknown holographic profilometry of the inner surface of a pipe using a
position and scan another section which overlaps with the current-induced wavelength change of a laser diode." Applied
former one; 3. Conduct point cloud registration between the optics 50.21 (2011): 3937-3946.
two section and unify them to a global coordinate system. The [3]Yokota, Masayuki, Teppei Koyama, and Tomoaki
measurement range could be extended by this scanning method. Kawakami. "Digital holographic inspection for the inner
Fig.13 shows an example of the registration result between surface of a straight pipe using current-induced
two adjacent sections. multiwavelength from two laser diodes." Optical Engineering
53.10 (2014): 104103-104103.
[4] Inari, Takahiko, et al. "Optical inspection system for the
inner surface of a pipe using detection of circular images
projected by a laser source." Measurement 13.2 (1994): 99-106.
[5] Zhuang B H, Zhang W, Cui D Y. Noncontact laser
sensor for pipe inner wall inspection [J]. Optical Engineering,
1998, 37(5): 1643-1647.
[6] Zhang W W, Zhuang B H. Non-contact laser inspection
for the inner wall surface of a pipe [J]. Measurement Science
and Technology, 1998, 9(9): 1380.
[7] Zhang G, He J, Li X. 3D vision inspection for internal
surface based on circle structured light [J]. Sensors and
Actuators A: Physical, 2005, 122(1): 68-75.
Fig.13 Point cloud registration result of two adjacent sections. [8] Wakayama T, Takano H, Yoshizawa T. Development of
But there are two remaining problems those are difficult to a compact inner profile measuring instrument [C]//Optics East
solve. The first one is the high time consumption due to huge 2007. International Society for Optics and Photonics, 2007:
data volume and computation, which might decrease if better 67620D-67620D-5.
algorithm is applied. The second one is that this method highly [9] Yoshizawa T, Wakayama T. Development of an inner
relies on feature points so that it is not appropriate for smooth profile measurement instrument using a ring beam device
surfaces e.g. air cylinders, which might limit the system’s [C]//Photonics Asia 2010. International Society for Optics and
application, although air cylinders are usually relatively short. Photonics, 2010: 78550B-78550B-8.
[10] Wakayama T, Machi K, Yoshizawa T. Small size probe
CONCLUSSIONS for inner profile measurement of pipes using optical fiber ring
A laser-triangulation-based 3D measurement system along beam device [C]//Photonics Asia. International Society for
with corresponding flexible calibration algorithm is developed Optics and Photonics, 2012: 85630L-85630L-7
in this research. Comparing to existing systems and methods, [11] Zhang, Zhengyou. "A flexible new technique for
this research takes refraction distortion and assembly errors into camera calibration." IEEE Transactions on pattern analysis and
consideration and compensates these errors through the machine intelligence 22.11 (2000): 1330-1334.
proposed binocular-vision-based calibration algorithm.

7 Copyright © 2018 ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/29/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like