USING CONVOLUTIONAL NEURAL NETWORK FOR IMAGE-BASED STEEL BRIDGE COATING RUST DETECTION VIA GOOGLE TENSORFLOW

簡易檢索 / 詳目顯示

回結果列表

研究生：	Kemo Jallow Kemo Jallow
論文名稱：	USING CONVOLUTIONAL NEURAL NETWORK FOR IMAGE-BASED STEEL BRIDGE COATING RUST DETECTION VIA GOOGLE TENSORFLOW USING CONVOLUTIONAL NEURAL NETWORK FOR IMAGE-BASED STEEL BRIDGE COATING RUST DETECTION VIA GOOGLE TENSORFLOW
指導教授：	謝佑明 Yo-Ming Hsieh 廖國偉 Kuo-Wei Liao
口試委員:	Yo-Ming Hsieh Yo-Ming Hsieh Kuo-Wei Liao Kuo-Wei Liao I-Tung Yang I-Tung Yang
學位類別：	碩士 Master
系所名稱：	工程學院 - 營建工程系 Department of Civil and Construction Engineering
論文出版年：	2018
畢業學年度：	106
語文別：	英文
論文頁數：	92
中文關鍵詞：	Steel Bridge 、Image-based 、Rust Detection 、Convolution 、Neural Network 、TensorFlow 、U-net
外文關鍵詞：	Rust Detection, Convolution
相關次數：	點閱：181 下載：6
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

Steel bridge rust inspection is unarguably one of most the important bridge inspection operations, it helps to schedule and determine which part of the steel bridge requires recoating. Inspection is often performed manually. However manual inspection is not efficient and reliable since it is based on human instincts and its evaluation result varies from one professional engineer to another. Alternatively, clustering such as K-means can be used to determine the percentage rusted area of the steel bridge, which is more consistent than visual inspection. However, clustering seems to yield an error when the bridge image pixels homogeneous.
This aim of this research is to use the Convolutional Neural Network (CNN) to determine rust percentage and evaluate its performance of different parts of the steel bridge. The machine learning library used in this research is Google TensorFlow. The CNN architecture adopted is U-net model. The U-net CNN model is trained on 80 images and tested on 20 images that are not visible to the model during the training. Then the accuracy rate and the rust percentages of the output images are computed. The average training and testing performances of the U-net CNN model were 78.9% and 76.8% respectively. American Society for Testing and Materials ( ASTM) was used to rate the corrosion performance of the steel bridge images. Each predicted image computed by the model is compared to its corresponding ground-truth which represent the true value of the input image.

ABSTRACT ..................................................................................................................I
LIST OF TABLES ...................................................................................................... V
LIST OF FIGURES....................................................................................................VI
NOTATIONS & TERMINOLOGIES ....................................................................... VIII
INTRODUCTION....................................................................................................1 1.1.
Background .....................................................................................................1 1.2.
Research Motivation........................................................................................3 1.3.
Research Objective..........................................................................................5 1.4.
Scope and Limitation....................................................................................... 5 1.5.
Thesis Outline..................................................................................................6
LITERATURE REVIEW ......................................................................................... 8
1. Clustering Based Approached ......................................................................... 8
2. Support Vector Machine (SVM) Approach ..................................................10
3. Artificial Neural Network (ANN) Approach ................................................ 10
U-NET RUST DETECTION METHOD ............................................................... 12
1. Toolboxes ...................................................................................................... 12
2. Image Data Processing .................................................................................. 12
2.1. Labeling Images.........................................................................................14
2.2. Resizing Images ......................................................................................... 14
3. Color Spaces Selection..................................................................................16
4. CNN Neural Network Architecture............................................................... 17
4.1. Input Layer.................................................................................................20
4.2. Convolutional Layer .................................................................................. 20
4.3. Activation Functions..................................................................................24
4.4. Pool ............................................................................................................ 26
4.5. Loss Function.............................................................................................28
5. Training Optimization ................................................................................... 31
5.1. Gradient Descent Optimization..................................................................33
III
5.2. Forward Propagation..................................................................................35
5.2. Backward Propagation ............................................................................... 41
RESULTS AND DISCUSSION ............................................................................43
1. Analysis Results ............................................................................................ 43
2. Performance Evaluation of the U-net CNN Model ....................................... 43
2.1. Training and Testing Accuracy Evaluation ............................................... 44
2.2. Image-Based Rust Percentage....................................................................44
CONCLUSION AND RECOMMENDATION ..................................................... 65
1. Summary and Conclusion ............................................................................. 65
2. Recommendations for Future Study..............................................................66
REFERENCES ........................................................................................................... 67
APPENDIX ................................................................................................................ 71
Software packages used ............................................................................... 71
Rust Detection Approach in TensorFlow ....................................................71
CNN Model Graph Nodes. .......................................................................... 76
                                

Abadi, Martín, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Ian Goodfellow, Andrew Harp, Geoffrey Irving, Michael Isard, Yangqing Jia, Rafal Jozefowicz, Lukasz Kaiser, Manjunath Kudlur, Josh Levenberg, Dan Mane, Rajat Monga, Sherry Moore, Derek Murray, Chris Olah, Mike Schuster, Jonathon Shlens, Benoit Steiner, Ilya Sutskever, Kunal Talwar, Paul Tucker, Vincent Vanhoucke, Vijay Vasudevan, Fernanda Viegas, Oriol Vinyals, Pete Warden, Martin Wattenberg, Martin Wicke, Yuan Yu, Xiaoqiang Zheng. TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems. www.tensorflow.org. (August 25, 2018).
Azizinamini, Power E Myers G Ozyildirim H Kline E Whitmore D Mertz D. 2013. Design Guide for Bridges for Service Life. Washington, D.C.: Transportation Research Board. https://www.nap.edu/catalog/22617 (August 25, 2018).
Barnett, Anthony Santokhi J Simpson M Smart N Stainton-Bygrave C Vivek S Waller A. Image Classification Using Non-Linear Support Vector Machines on Encrypted Data. https://eprint.iacr.org/2017/857.pdf (August 25, 2018).
Chen, Jieh Haur, Su M Cao R Hsu S Lu J. 2017. “A Self Organizing Map Optimization Based Image Recognition and Processing Model for Bridge Crack Inspection.” Automation in Construction 73: 58–66. http://dx.doi.org/10.1016/j.autcon.2016.08.033.
Chen, Po Han, H K Shen, C Y Lei, and L M Chang D A Associate. 2011. “Fourier- Transform-Based Method for Automated Steel Bridge Coating Defect Recognition Selection and/or Peer-Review under Responsibility of [Name Organizer] Open Access under CC BY-NC-ND License. Open Access under CC BY-NC-ND License.” www.sciencedirect.com (August 25, 2018).
Chen, Po-han, Heng-kuang Shen, and Luh-maan Chang. “A RUST INTENSITY RECOGNITION APPROACH FOR STEEL BRIDGE SURFACE COATING IMAGES BASED ON ARTIFICIAL NEURAL.” : 996–1001.
Chen, Po-Han, Ya-Ching Yang, and Luh-Maan Chang. 2010. “Box-and-Ellipse-Based ANFIS for Bridge Coating Assessment.” Journal of Computing in Civil
67
Engineering 24(5): 389–98. http://search.ebscohost.com/login.aspx?direct=true&db=a9h&AN=52929246&si te=ehost-live.
Chen, Po Han, Heng Kuang Shen, Chi Yang Lei, and Luh Maan Chang. 2012. “Support-Vector-Machine-Based Method for Automated Steel Bridge Rust Assessment.” Automation in Construction 23: 9–19. http://dx.doi.org/10.1016/j.autcon.2011.12.001.
Chen, Po Han, Ya Ching Yang, and Luh Maan Chang. 2009. “Automated Bridge Coating Defect Recognition Using Adaptive Ellipse Approach.” Automation in Construction 18(5): 632–43. http://dx.doi.org/10.1016/j.autcon.2008.12.007.
Chen, Yushi, Jiang H, Li C Jia X Ghamisi P. 2016. “Deep Feature Extraction and Classification of Hyperspectral Images Based on Convolutional Neural Networks.” IEEE Transactions on Geoscience and Remote Sensing 54(10): 6232– 51. http://ieeexplore.ieee.org/document/7514991/ (August 27, 2018).
DAHAL, PARAS. 2016. “Classification and Loss Evaluation - Softmax and Cross Entropy Loss.” https://deepnotes.io/softmax-crossentropy (August 28, 2018).
Dhanachandra, Nameirakpam, Khumanthem Manglem, and Yambem Jina Chanu. 2015. “Image Segmentation Using K-Means Clustering Algorithm and Subtractive Clustering Algorithm.” Procedia Computer Science 54(December): 764–71. http://dx.doi.org/10.1016/j.procs.2015.06.090.
DiPietro, Rob. 2016. “A Friendly Introduction to Cross-Entropy Loss.” https://rdipietro.github.io/friendly-intro-to-cross-entropy-loss/ (August 28, 2018).
Dumoulin, Vincent, Francesco Visin, and George E P Box. 2018. A Guide to Convolution Arithmetic for Deep Learning. http://ethanschoonover.com/solarized (August 27, 2018).
Fletcher, Tristan. 2008. Support Vector Machines Explained. www.cs.ucl.ac.uk/staff/T.Fletcher/ (August 27, 2018).
Florida, A, and West Lafayette. 2000. “I m c f s A.” 126(December): 422–32.
Golhani, Kamlesh, Siva K. Balasundram, Ganesan Vadamalai, and Biswajeet Pradhan. 2018. “A Review of Neural Networks in Plant Disease Detection Using Hyperspectral Data.” Information Processing in Agriculture 5(3): 354–71.
68

https://linkinghub.elsevier.com/retrieve/pii/S2214317317301774 (August 27, 2018).
Golik, P, and P Doetsch. 2013. “Cross-Entropy vs. Squared Error Training:A Theoretical and Experimental Comparison.” Interspeech, Isca 2(2): 1756–60.
Gu, Jiuxiang Wang, Z Kuen, J Ma, L Shahroudy, A Shuai, B Liu, T Wang, X Wang, L Wang, G Cai, J Chen T. 2015. “Recent Advances in Convolutional Neural Networks.” : 1–38. http://arxiv.org/abs/1512.07108.
isaac changhau. 2017. “Loss Functions in Neural Networks | Isaac Changhau.” https://isaacchanghau.github.io/post/loss_functions/ (August 27, 2018).
Kim, Changwan, Hyojoo Son, Nahyae Hwang, and Changmin Kim. 2014. “Rapid and Automated Determination of Rusted Surface Areas of a Steel Bridge for Robotic Maintenance Systems.” Automation in Construction 42: 13–24. http://dx.doi.org/10.1016/j.autcon.2014.02.016.
Kingma, Diederik P, and Jimmy Lei Ba. 2017. ADAM: A METHOD FOR STOCHASTIC OPTIMIZATION. https://arxiv.org/pdf/1412.6980.pdf (August 28, 2018).
Krause, Andreas, and Nathan Watson. 2010. CS/CNS/EE 253: Advanced Topics in Machine Learning Topic: Nonparametric Learning and Gaussian Processes Lecturer: From the Last Lecture, We Have the Following General Formulation for Learning Problems: F * = Min. http://courses.cms.caltech.edu/cs253/slides/cs253-14-GPs.pdf (August 28, 2018).
Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E Hinton. 2012. “ImageNet Classification with Deep Convolutional Neural Networks.” Advances In Neural Information Processing Systems: 1–9.
Liao, Kuo Wei, and Yi Ting Lee. 2016. “Detection of Rust Defects on Steel Bridge Coatings via Digital Image Recognition.” Automation in Construction 71(Part 2): 294–306. http://dx.doi.org/10.1016/j.autcon.2016.08.008.
ML Cheatsheet. 2017. “Loss Functions — ML Cheatsheet Documentation.” https://ml- cheatsheet.readthedocs.io/en/latest/loss_functions.html (August 28, 2018).
Nair, Vinod, and Geoffrey E Hinton. Rectified Linear Units Improve Restricted Boltzmann Machines.
69

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.165.6419&rep=rep1& type=pdf (August 27, 2018).
Narayan, Sridhar. 1997. The Generalized Sigmoid Activation Function: Competitive Supervised Learning. https://ac.els-cdn.com/S0020025596002009/1-s2.0- S0020025596002009-main.pdf?_tid=eac4a9f6-a0fa-4c6a-b497- 193912e4faa2&acdnat=1535386865_ece3aa991b896bde022ddf235c706cdc (August 28, 2018).
Qiu, Suo, Xiangmin Xu, and Bolun Cai. 2018. FReLU: Flexible Rectified Linear Units for Improving Convolutional Neural Networks. https://github.com/facebook/fb.resnet.torch (August 27, 2018).
Rawat, Waseem, and Zenghui Wang. 2017. “Deep Convolutional Neural Networks for Image Classification: A Comprehensive Review.” Neural computation 29(9): 2352–2449. http://www.ncbi.nlm.nih.gov/pubmed/28599112 (August 28, 2018).
———. 2018. “Deep Convolutional Neural Networks for Image Classification: A Comprehensive Review.” 2449: 2352–2449. http://arxiv.org/abs/1803.01446.
Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. 2015. U-net: Convolutional Networks for Biomedical Image Segmentation. http://lmb.informatik.uni- freiburg.de/ (August 25, 2018).
Ruder, Sebastian. 2017. An Overview of Gradient Descent Optimization Algorithms *. http://caffe.berkeleyvision.org/tutorial/solver.html (August 28, 2018).
Shen, Heng Kuang, Po Han Chen, and Luh Maan Chang. 2013. “Automated Steel Bridge Coating Rust Defect Recognition Method Based on Color and Texture Feature.” Automation in Construction 31: 338–56. http://dx.doi.org/10.1016/j.autcon.2012.11.003.
Troitsky, Michael Serge. 1994. “Planning and Design of Bridges.” In 2nd Edition, J. Wiley, 109–12. https://books.google.com.tw/books?hl=en&lr=&id=v96pMiD4A9UC&oi=fnd&p g=PR17&ots=S6vkRj7j2v&sig=P5iyJrvZWpqR6yCa5j7ImGVskAQ&redir_esc =y#v=onepage&q&f=false (August 25, 2018).

簡易檢索 / 詳目顯示

相關論文