Professional Documents
Culture Documents
Carpi 2010
Carpi 2010
net/publication/224190295
CITATIONS READS
33 1,930
4 authors, including:
Some of the authors of this publication are also working on these related projects:
Automated Flooded area detection using remotely sensed images View project
Vision based aircraft detection for UAS detect and avoid View project
All content following this page was uploaded by Luis Mejias on 30 May 2014.
T REES, shrubs and other vegetation are of continued a stereo vision system [4], and unmanned aerial systems
importance to the environment and our daily life. They (UASs) [5, 12]. Therefore, remotely sensed data have great
provide shade around our roads and houses, offer a habitat potential in assisting vegetation management in power line
for birds and wildlife, and absorb air pollutants. However, corridors.
vegetation around power lines can represent a significant risk This work is closely aligned with industry, with strong
to public safety, the environment and, is one of the main support and collaboration from Ergon Energy. In this paper,
causes of power outages. It is common knowledge that trees we present a comprehensive discussion of vegetation
falling across power lines are the largest cause of power management in power line corridor monitoring based on aerial
failures, causing widespread power outages and bushfires [1]. remote sensing techniques. We focus on discussing the risk
It is not surprising that vegetation management procedures in management strategy, selection of platform and sensor,
power line corridors become a significant maintenance cost advancement in automated data processing, and the challenges
for electrical companies. Currently, activities such as tree learnt from our project.
trimming and vegetation control are costly. For example,
Ergon Energy, one of the Australia’s largest electricity II. BACKGROUND
companies, currently spends $80 million a year inspecting and
managing vegetation encroachments on power lines. Correct This section provides a summary of our research covering
and efficient vegetation management not only reduces the the risk management strategy using aerial remote sensing
overall cost but also aids in continuous electricity supply by techniques, selection of platform and sensor, and automated
preventing damage to power lines through removal of tall- feature extraction from remotely sensed data.
growing species. Ineffective procedures can result in the loss A. Risk Management Strategy
of reliability in electricity transmission, produce serious
The reliability of electricity supply and distribution is of the
This work was conducted within the CRC for Spatial Information,
highest priority in power line corridor monitoring. A short-
established and supported under the Australian Government’s Cooperative term strategy is to identify and remove nearby objects (i.e.
Research Centres Programme, and sees collaboration between the Queensland buildings and vegetation) around power lines. Generally, the
University of Technology (QUT), the Australian Research Centre for
Aerospace Automation (ARCAA) and Ergon Energy Australia.
risk of manmade structures can be controlled through building
Zhengrong Li and Ross Hayward are with the Faculty of Science and regulations. However, vegetation is naturally growing and
Technology, Queensland University of Technology (QUT), Brisbane, QLD, particularly in rural areas the growth is unmanaged. Strong
4001, Australia (e-mail: zhengrong.li@qut.edu.au; r.hayward@qut.edu.au). winds and storms can bring branches or even entire trees into
Rodney Walker and Luis Mejias are with the Australian Research Centre
for Aerospace Automation (ARCAA, QUT), Brisbane, QLD, 4009, Australia contact with power lines. Unmanaged vegetation can also
(ra.walker@qut.edu.au; luis.mejias@qut.edu.au).
grow up into power lines and cause bushfires. These problems heavy payloads. LiDAR systems are too heavy for
motivate the use of aerial remote sensing techniques for power small/medium sized UAS platforms. This limitation may be
line corridor monitoring. Specifically, detecting all vegetation overcome in the near future as there are already small LiDAR
from remote sensed data that have the potential to pose risks to systems in the market suitable for UASs. However, the
power lines and use this to guide the field workers for performance of these units in terms of the quality of data
vegetation clearance in the corridors. collected is currently well away from their full-sized
Ergon Energy has a long-time strategy of managing counterparts. When combined with LiDAR type systems
vegetation according to different species. The species they are UASs would represent the technology of choice for future uses.
interested in can be generally categorized as undesirable
C. Automated Data Processing
species and undesirable species. Species with fast growth rates
and that also have the potential to reach a mature height of As may already be evident from previous section, aerial
more than four meters are defined as undesirable species. remote sensing is a fast and cost-effective technique for power
These undesirable species often pose high risks to electrical line corridor mapping. There are some commercial data
infrastructure and therefore should be identified and removed. providers available in the market that can collect high quality
It is also worth mentioning that in the long-term maintenance LiDAR data and georeferenced imagery for general mapping
strategy low-growing trees or shrubs are encouraged because purposes. Compared to corridor mapping, automated and
they are expected to compete with tall growing species and intelligent information extraction from remotely sense data is
deprive the taller trees of light and nutrients. These low more challenging. One special need for power line corridor
growing species, along with the rare and endangered species, monitoring is to detect the objects of interest for further
are defined as desirable species that should be managed interpretation and decision making. The major objects of
differently. interest include power line assets and vegetation. Automated
data processing aims to automatically detect these objects
B. Selection of Platform and Sensor from aerial imagery, and try to extract more specific
Satellites and aircrafts are the most widely used platforms information such as vegetation species and height information.
for remote sensing in earth observing data collection. Current In this section, we briefly discuss the general methods in
satellite sensors are not the best choice for monitoring power automated data processing.
line corridors due to two critical limitations: the unfavorable • Power Line Detection
revisit time and lack of choices in optimum spatial and Risk assessment of power lines and adjacent trees is
spectral resolutions. At the most practical level, most meaningful only when power lines can be detected. A straight
collections of data gathered from satellites are available only line can approximate a power line segment in aerial images.
on predetermined schedules, and even those with an “on- Therefore, classic line detection algorithms like the Hough
demand” capability are also limited by their orbits and the transform may be used to detect power lines in images.
demands of other users. In contrast, airborne data collection However, while the Hough transform is an effective tool to
offers a much greater level of flexibility. An airborne system detect straight lines, it does not intelligently identify power
can capture data at any time of the day whereas satellites lines. Any linear objects will be detected, such as fences,
generally pass over one site at the same time of a day. Another edges of roads and buildings. Therefore, some prior
advantage of airborne platform is that different sensor knowledge should be used to improve power line detection [5].
payloads can be easily fitted, while the sensors launched on a For 3D LiDAR point cloud data, line detection can be
satellite are often not changeable. As a consequence, airborne conducted either by clustering similar features in a voxel [13]
systems can be regularly upgraded as sensor technology or mapping 3D data to two 2D planes (horizontal and vertical)
advances. Improvements to sensors include systems with and then roughly detecting the power line points in horizontal
higher spectral and spatial resolution, and advanced plane and reconstructing the catenary curve in the vertical
microwave or LiDAR sensors. In addition, higher spatial plane [14]. No matter which method is used, prior removal of
resolutions are easier to obtain from airborne platforms, due to non-powerline points (e.g. terrain, tree and building) will
their low altitude. A limitation which impedes large-scale always be helpful to reduce the error in power line detection.
airborne remote sensing applications is that the traditional • Tree Detection and Classification
piloted airborne platforms involve high operational costs. Detection of trees has been intensively studied particularly
Moreover, using piloted aircraft for power line inspection will in the remote sensing of forest environment. Whilst similar in
place the operators at a greater level of risk. concept, the environment in power line corridor is more
Remote sensors mounted on unmanned aerial systems complex because the background is cluttered with shadows,
(UASs) could fill this gap, providing a cheap and flexible way bare soil, shrubs and grass; all presenting irregularities that
to gather spatial data from power line corridors which can also need to be handled by the detection algorithm. Vegetation has
meet the requirements of spatial, spectral, and temporal a distinctive spectral signature, characterized by a low
resolutions. Recent development in the aerial vehicles reflectance in the visible part of the solar spectrum, and a high
themselves and associated sensing system make UAS reflectance in the near-infrared (NIR) region. Therefore,
platforms increasingly attractive for both research and incorporating NIR information is widely used in the remote
operational mapping [6, 7]. One of the main limitations of sensing community for the detection and classification of
using UASs is their ability to carry power-demanding and
3
vegetation. Combining LiDAR elevation data can further species in power line corridors requires significantly more
improve tree detection by removing low-growing grass and resources than are currently available, however, classifying
shrubs. For classification of tree species, object-based methods species in a given test area as a proof of concept is possible. In
are preferred as they are straightforward and have been shown this research, we focus on three dominant species in our test
to obtain higher classification accuracy in high resolution field: Eucalyptus tereticornis, Eucalyptus melanophloia, and
image classification [8]. To conduct an object-based Corymbia tesselaris. We abbreviate the species names to
classification, accurate individual tree segmentation is Euc_Ter, Euc_Mel and Cor_Tes. For each species 65 samples
required. Subsequent to this, a range of classification were used with 25 for training and 40 for testing.
algorithms can be used in the object-feature space.
(a) power line detection from aerial imagery (a) original image
VIII. BIOGRAPHIES
Zhengrong Li received the B. Eng. and M. Eng.
degree in 2003 and 2006 respectively from
Northwest A&F University, China. He was
appointed an assistant lecturer at the College of
Information Engineering, Northwest A&F University
since 2006. He is currently working toward the Ph.D.
degree at the Queensland University of Technology
with the support of Cooperative Research Centre for
Spatial Information (CRCSI). His research interests
include remote sensing image processing,
computational intelligence, and pattern recognition.