The Complete Material is Available. View Abstract or Chapter One Below.

Download this complete Project material titled; Development Of An Improved Algorithm For Power Line Detection In Optical Images Using Frangi Filter And First Order Derivative Of Gaussian with abstract, chapters 1-5, references, and questionnaire. Preview Abstract or chapter one below

  • Format: PDF and MS Word (DOC)
  • pages = 65

 3,000

100% Money-Back Guarantee

Do you need help?

Call or Whats-app us: (+234) 08060082010, 08107932631.

ABSTRACT

 

This research presents the development of a Frangi filter and first-order derivative of Gaussian (FF-FDOG) based power line detection (PLD) algorithm as an improvement to the standard PLD algorithm. Vision-based PLD is important in obstacle avoidance in low-altitude flight and also in the surveillance and maintenance of electrical infrastructure. The need for high and real-time detection rates as well as low false alarm in noisy and cluttered images makes it a challenging task. Matched filterand first-order derivative of Gaussian (MF-FDOG) based PLD algorithmwas developed to handle limitations associated with the standard PLD algorithm in terms of its ability to automatically select a problem specific threshold in its edge detection. The MF-FDOG based PLD, however, returned a high false positive rate and a detection rate insufficient for real time processing.The FF-FDOG based threshold is developed in this work using frangi filter (which detects vessel based on the eigen value analysis of the second order structure of an image) and FDOG filter. Images from the University of South Florida computer vision and pattern recognition group wire database were used to evaluate the performance of the developed FF-FDOG method. In the results obtained, the true positive rate of the developed FF-FDOG based PLD algorithm was 86.39%, which is a 2.64% improvement over MF-FDOG’s 84.16%, while the false positive rate of the developed FF-FDOG based PLD algorithm was 11.45%, which is a 36.06% improvement over MF-FDOG’s 17.91%.

 

TABLE OF CONTENTS

DECLARATION ……………………………………………………………………………………………………….. ii
CERTIFICATION ……………………………………………………………………………………………………. iii
DEDICATION ………………………………………………………………………………………………………….. iv
ACKNOWLEDGEMENT …………………………………………………………………………………………… v
ABSTRACT ……………………………………………………………………………………………………………… vi
TABLE OF CONTENTS …………………………………………………………………………………………… vii
LIST OF FIGURES ……………………………………………………………………………………………………. x
LIST OF PLATES …………………………………………………………………………………………………….. xi
LIST OF TABLES ………………………………………………………………………………………………….. xiii
ABBREVIATIONS …………………………………………………………………………………………………. xiv
CHAPTER ONE ………………………………………………………………………………………………………… 1
INTRODUCTION ……………………………………………………………………………………………………… 1
1.1 Background ………………………………………………………………………………………………….. 1
1.2 Statement of the Problem ……………………………………………………………………………….. 3
1.3 Significance of the Study ……………………………………………………………………………….. 3
1.4 Aim and Objectives of the Study …………………………………………………………………….. 3
1.5 Methodology ………………………………………………………………………………………………… 4
1.6 Dissertation Organization ……………………………………………………………………………….. 5
CHAPTER TWO ……………………………………………………………………………………………………….. 7
LITERATURE REVIEW ……………………………………………………………………………………………. 7
2.1 Introduction ………………………………………………………………………………………………….. 7
2.2 Review of Fundamental Concepts……………………………………………………………………. 7
2.2.1 Power line detection ………………………………………………………………………………… 7
2.2.2 Power line detection algorithm …………………………………………………………………. 9
2.2.3 MF-FDOG power line segment detection…………………………………………………. 10
2.2.4 Multiscale vessel enhancement filter: Frangi filter …………………………………….. 15
2.2.5 Graph-cut model based line detection………………………………………………………. 22
viii
2.2.6 Morphological operations ………………………………………………………………………. 25
2.2.7 Performance metrics ……………………………………………………………………………… 28
2.3 Review of Similar Works ……………………………………………………………………………… 30
CHAPTER THREE ………………………………………………………………………………………………….. 37
MATERIALS AND METHODS ………………………………………………………………………………… 37
3.1 Introduction ………………………………………………………………………………………………… 37
3.2 Experimental data. ……………………………………………………………………………………….. 37
3.2.1 Image Acquisition. ………………………………………………………………………………… 37
3.2.2 Clutter measure …………………………………………………………………………………….. 40
3.2.3 Parameters ……………………………………………………………………………………………. 41
3.3 Replication of the MF-FDOG based PLD algorithm ………………………………………… 42
3.3.1 Pre-processing the Image ……………………………………………………………………….. 42
3.3.2 Matched filtering …………………………………………………………………………………… 42
3.3.3 First Order derivative of Gaussian filtering ………………………………………………. 43
3.3.4 MF-FDOG thresholding…………………………………………………………………………. 45
3.3.5 Morphological filtering ………………………………………………………………………….. 45
3.3.6 Line segment clustering …………………………………………………………………………. 46
3.4 Development of the FF-FDOG based PLD algorithm ………………………………………. 47
3.4.1 Frangi filtering ……………………………………………………………………………………… 48
3.4.2 FF-FDOG thresholding ………………………………………………………………………….. 49
3.5 Performance measures………………………………………………………………………………….. 50
CHAPTER FOUR …………………………………………………………………………………………………….. 52
RESULTS AND DISCUSSION …………………………………………………………………………………. 52
4.1 Introduction ………………………………………………………………………………………………… 52
4.2 Clutter Measure …………………………………………………………………………………………… 52
4.3 Experimental Results……………………………………………………………………………………. 53
4.3.1 Visualized results ………………………………………………………………………………….. 53
4.3.2 Quantified results ………………………………………………………………………………….. 64
4.4 Discussion ………………………………………………………………………………………………….. 65
4.5 Application of FF-FDOG on captured images …………………………………………………. 65
ix
CHAPTER FIVE ……………………………………………………………………………………………………… 70
SUMMARY, CONCLUSION AND RECOMMENDATIONS ………………………………………. 70
5.1 Summary ……………………………………………………………………………………………………. 70
5.2 Conclusion ………………………………………………………………………………………………….. 70
5.3 Significant Contributions ……………………………………………………………………………… 71
5.4 Limitations …………………………………………………………………………………………………. 71
5.5 Recommendations for Further Work………………………………………………………………. 71
REFERENCES ………………………………………………………………………………………………………… 72
APPENDIX A1 ………………………………………………………………………………………………………… 77
MATLAB code for MF-FDOG PLD algorithm ……………………………………………………………. 77
APPENDIX A2 ………………………………………………………………………………………………………… 81
MATLAB code for FF-FDOG PLD algorithm ……………………………………………………………… 81
APPENDIX B1 ………………………………………………………………………………………………………… 85
MATLAB code for matched filtering ………………………………………………………………………….. 85
APPENDIX B2 ………………………………………………………………………………………………………… 86
MATLAB code for frangi filtering ……………………………………………………………………………… 86
APPENDIX B3 ………………………………………………………………………………………………………… 89
MATLAB code for first order derivative of Gaussian filtering ……………………………………….. 89
APPENDIX C1 ………………………………………………………………………………………………………… 90
MATLAB code for performance measure ……………………………………………………………………. 90
APPENDIX C2 ………………………………………………………………………………………………………… 92
Data for MF-FDOG performance measure …………………………………………………………………… 92
APPENDIX C3 ………………………………………………………………………………………………………… 94
Data for FF-FDOG performance measure ……………………………………………………………………. 94
APPENDIX D1 ………………………………………………………………………………………………………… 96
MATLAB code for clutter measure …………………………………………………………………………….. 96
APPENDIX D2 ………………………………………………………………………………………………………… 97
Data for clutter measure …………………………………………………………………………………………….. 97

 

Project Topics

 

CHAPTER ONE

INTRODUCTION
1.1 Background
Electricity is vital for the activities of modern-day societies and effective monitoring and maintenance of power lines are needed to secure uninterrupted distribution of electricity, (Matikainen et al., 2016). Electricity companies spend a significant budget on power line inspections, and continuously pursue new approaches to reduce inspection cost (Martinez et al., 2014). For example, Ergon Energy, one of the top electricity companies in Australia, spends $80 million a year inspecting and managing vegetation that encroaches on power line assets (Li et al., 2010). Inspection of high voltage power lines can be very dangerous if performed by humans, very expensive if performed by helicopters and damaging on the cable if performed by the roll robot. Unmanned aerial vehicle (UAV) is, therefore, one of the best instruments for detailed power line inspection tasks (Zhou et al., 2016).
Thin objects such as cables, power lines, and wires (whose position cannot be guaranteed as known before flight) are not easily perceived by helicopters and small UAV pilots over heavily cluttered backgrounds, or when the contrast between the object and the background is low (Byrne et al., 2006; Candamo et al., 2009; Luo et al., 2014). Power line detection is a widely sought environmental awareness techniques and it is of great significance in ensuring the safety of low altitude flight which is essential in a variety of applications (such as traffic and power infrastructure monitoring, border patrol, search and rescue, and surveillance), as even a low-speed collision with power lines can be fatal (Gandhi et al., 2003; McGee et al., 2005; Scherer et al., 2008).
In particular, comprehensive real data suggests that more than half of the low altitude aerial collision accidents are caused by power lines(Shan et al., 2015). The United State (U.S.)
2
Army reports that more helicopters have been lost due to hitting power lines than are to combat(Avizonis & Barron, 1999). Between 1997 and 2006 the U.S. Army recorded 54 so-called “wire strikes” in which 13 military personnel died and which caused $224 million in damages. Also in the same period there were 102 civilian wire strikes, killing 33 people, according to data from the National Transportation Safety Board (NTSB)(Safe Flight, 2009).
All these highlightthe need forenhanced capabilities in power line detection(PLD) techniques that should provide a high probability of timely detection while maintaining a low probability of false alarm in noisy, cluttered images of power lines exhibiting a wide range of sizes and complexities (Kasturi & Camps, 2002).
Continued advances in the fields of image-processing and computer vision have raised interest in their suitability in PLD for collision avoidance and also for power line tracking during inspection with UAVs (Gandhi et al., 2003; Seibold et al., 2013). Optical based power line detection algorithms generally capture two criteria(local and global criterion)in which a threshold is set manually for its edge detection in the local criterion (Bhujade et al., 2013; Cao et al., 2013; Liu et al., 2012; Seibold et al., 2013; Yang et al., 2012; Zhang et al., 2012; Zhu et al., 2013). However, a fixed threshold are only effective when the background is relatively monotonic thus they may cause these algorithms to fail with complex and changing backgrounds (Zhou et al., 2016).A problem-specific threshold schemewas developed bySong & Li (2014),to overly detect power line segments which was a variant of a retinal blood vessel detection method based on the extension of matched filter (MF), proposed by Zhang et al. (2010). The thresholding scheme was able to detect symmetrical edges and suppress step edges in images of powerline but it still produced a high false alarm.
To this end, Frangi filter (FF), a vessel enhancement filter, in which vesselness measure is obtained on the basis of all eigenvalues of the multiscale second order local structure of an
3
image (Frangi et al., 1998), and first order derivative of Gaussian(FDOG) is employed in the local criterion of the PLD algorithm in this research.
1.2 Statement of the Problem
Optical based power line detection algorithms generally capture two criteria (local and global criterion) in which a threshold is set manually for its edge detection in the local criterion. However, a fixed threshold are only effective when the background is relatively monotonic thus they may cause these algorithms to fail with complex and changing backgrounds. Anadaptive threshold schemebased on the image response to MF and FDOG was introduced to mitigate the problem of fixed threshold. However, it is not without its own drawbacks as it returned a high false positive rate and a detection rate insufficient for real time processing.This research is aimed at developing an adaptive threshold scheme based on the image response to FF and FDOG which will maximise the true positive detection rate and minimise the false positive detection rate and runtime.
1.3 Significance of the Study
Power line detection is a widely sought environmental awareness techniques which reduce significantly, the budget Electricity companies spend on power line inspections, and the occurrence of “wire strikes” in low altitude flight, thus preventing loss of lives and properties. This Study offers a FF and FDOG based power line detection algorithm whichprovides a high probability of timely detection while maintaining a low probability of false alarm in noisy, cluttered images of power lines exhibiting a wide range of sizes and complexities
1.4 Aim and Objectives of the Study
The aim of this research work is the development of an improvedalgorithm for PLD in optical images using the Frangi filter and first order derivative of Gaussian (FF-FDOG). To achieve the stated aim, a number of objectives have been identified as follows:
4
1. To replicate the standard matched filter and first order derivative of Gaussian (MF-FDOG)and develop the FF-FDOG based power line detection algorithms
2. To compare the performance of the FF-FDOG and MF-FDOG based PLD algorithmsusing true positive rate, false positive rate and run time as metrics using the work of Song & Li (2014).
3. To apply the FF-FDOG based PLD algorithm on select dataset of 10 images of power lines obtained under different conditions from Samaru, Sabon-Gari Local Government Area of Kaduna State.
1.5 Methodology
The methodology adopted in this research is described as follows:
1. Replication and implementation of the MF-FDOG based PLD algorithm by adopting the following:
a) An edge map based on MF and FDOG is applied to detect all the line segments with symmetrical edges in the image.
b) A morphological filter is designed specifically to filter out the non-power line candidates.
c) The graph-cut model based on the graph theory is exploited to group the line segment pool into whole line pool
d) The “true” power lines is picked-up by morphology properties.
2. Development and implementation of the FF-FDOG based PLD algorithm by adopting the following:
a) An edge map based on FF and FDOG is applied to detect all the line segments with symmetrical edges in the image
b) Step 1 (b – d) is repeated.
5
3. Validation by comparing the performance of both algorithm by adopting the
following:
a) A dataset of 50 images of power lines is obtained from the University of South
Florida, computer vision and pattern recognition wire database.
b) For each image, clutter measure on the visual complexity of the background is
calculated to quantitatively show the richness of the dataset
c) For each power line, the ground truth is obtained by first manually labelling
several points, and then approximating with a straight line or a quadratic
polynomial.
d) The performance of both algorithms are measured by true positive rate (TPR) and
false positive rate (FPR) and run-time at line-level.
e) Comparison based on TPR, FPR and run time of both algorithms.
4. Application of the FF-FDOG based PLD algorithm on selected images under
different conditions by adopting the following:
a) A dataset of 10 images of power lines having resolution of 720 480 pixels, with
different noise level, ambiguity, weather conditions, backgrounds, is obtained
from Samaru, Sabon-Gari Local Government Area of Kaduna State using a
Digital Camera.
b) The FF-FDOG based PLD algorithms is applied on the images.
1.6 Dissertation Organization
The general introduction has been presented in Chapter One. The rest of the chapters are
structured as follows: Firstly, detailed review of related literature and relevant fundamental
concepts ofPLD, MF-FDOG based PLD algorithm,multiscale vessel enhancement
filter,graph-cut model based line detection, and morphological operations are discussed in
Chapter Two. Secondly, an in-depth approach and relevant mathematical models describing
6
the acquired image data, the replication of the MF-FDOG based PLD algorithm and the development of the FF-FDOG based PLD algorithm are presented in Chapter Three. Next, the analysis, performance and discussion of the result are shown in Chapter Four. Finally, summary, conclusion and recommendations for further work are presented Chapter Five. The list of cited references and MATLAB codes are provided in the appendices at the end of this dissertation.

GET THE COMPLETE PROJECT»
Do you need help? Talk to us right now: (+234) 08060082010, 08107932631, 08157509410 (Call/WhatsApp). Email: edustoreng@gmail.com