铁轨图像的低秩矩阵分解缺陷检测
|
摘要
随着高铁在中国乃至世界的快速发展,对轨道质量的要求越来越高。轨道表面缺陷检测直接关系着铁路安全、国家经济及安全等问题。基于固定光源下的摄像头拍摄的轨道表面图像,将轨道表面的缺陷检测建模为低秩矩阵分解问题,并对分解得到的稀疏矩阵计算其行累积量,由于去除了背景的干扰,缺陷区域所在的行累积量绝对值会变得很小,从而可以通过阈值操作得到缺陷行坐标,最后对由缺陷行坐标确定的小图像块区域进行二值化操作,并寻找最大联通区域即可确定缺陷位置,实现缺陷的自动检测与定位。与已有文献实验比较结果表明,该算法对不同光照及背景下的轨道缺陷检测均取得了较好的检测结果。
With the rapid development of high speed railway in China and over the world, the quality requirements of rail tracks become higher and higher. Rail track quality is directly related to the train safety, national economy, security and others. Based on rail track surface images obtained under fixed light sources, we transfer the rail track surface defect detection problem into a low-rank matrix decomposition problem. For the obtained sparse matrix, the accumulation value of each row is calculated. Because the background is removed in the sparse matrix, the absolute row accumulation values of the defect area will be very small. Thus a threshold operation can be applied and the row indexes of the defect areas can be obtained. Finally, binarization operation is applied to the small image block determined by the row indexes and the maximal connected region in the binarization image block can be considered as the defect area. Our algorithm can realize the automatic defect detection and localization. Compared with the exit algorithms, our algorithm achieve a better results under varied illumination and backgrounds.
With the rapid development of high speed railway in China and over the world, the quality requirements of rail tracks become higher and higher. Rail track quality is directly related to the train safety, national economy, security and others. Based on rail track surface images obtained under fixed light sources, we transfer the rail track surface defect detection problem into a low-rank matrix decomposition problem. For the obtained sparse matrix, the accumulation value of each row is calculated. Because the background is removed in the sparse matrix, the absolute row accumulation values of the defect area will be very small. Thus a threshold operation can be applied and the row indexes of the defect areas can be obtained. Finally, binarization operation is applied to the small image block determined by the row indexes and the maximal connected region in the binarization image block can be considered as the defect area. Our algorithm can realize the automatic defect detection and localization. Compared with the exit algorithms, our algorithm achieve a better results under varied illumination and backgrounds.
引文
[1] 许鹏,朱晨露,徐中行,等.基于差分涡流检测的铁轨裂纹特征识别方法[J].无损检测,2018,40(12):7-11.Xu Peng,Zhu Chenlu,Xu Zhongxing,et al.Rail Crack Identification Method based on Differential Eddy Current Testing[J].Nondestructive Detection,2018,40(12):7-11.(in Chinese)
[2] 戴鹏,王胜春,杜馨瑜,等.基于半监督深度学习的无砟轨道扣件缺陷图像识别方法[J].中国铁道科学,2018,39(4):45-51.Dai Peng,Wang Shengchun,Du Xinyu,et al.Image recognition Method for the Fastener Defect of Ballastless Track based on Semi-Supervised Deep Learning[J].China Railway Science,2018,39(4):45-51.(in Chinese)
[3] 刘孟轲,吴洋,王逊.基于卷积神经网络的轨道表面缺陷检测技术实现[J].现代计算机,2017,10(中):65- 69.Liu Mengke,Wu Yang,Wang Xun.Implementation of Track Surface Defect Inspection Based on Convolutional Neural Network[J].Modern Computer,2017,10(Middle):65- 69.(in Chinese)
[4] 茅正冲,邬锋.基于图像传感器的铁轨表面缺陷视觉检测算法[J].传感器与微系统,2015,34(9):141-144.Mao Zhengchong,Wu Feng.Visual Detection Algorithm for Rail Surface Defects based on Image Sensor[J].Transducer and Microsystem Technologies,2015,34(9):141-144.(in Chinese)
[5] 刘雄祥.基于卷积神经网络的铁轨表面缺陷识别研究[D].西南科技大学硕士论文,2018.Liu Xiongxiang.Research on Rail Surface Defect Recognition based on Convolutional Neural Network[D].Thesis for Master Degree of the Southwest University of Science and Technology,2018.(in Chinese)
[6] 宋亚帆,潘迪夫,韩锟.复杂环境下基于视觉显著性特征的铁轨识别方法[J].铁道科学与工程学报,2018,15(4):57- 65.Song Yafan,Pan Difu,Han Kun.Track Detection Approach in Complex Environment based on Saliency Features[J].Journal of Railway Science and Engineering,2018,15(4):57- 65.(in Chinese)
[7] Gan Jinrui,Li Qingyong,Wang Jianzhu,et al.A Hierarchical Extractor-Based Visual Rail Surface Inspection System[J].IEEE Sensors Journal,2017,17(23):7935-7944.
[8] Li Qingyong,Ren Shengwei.A visual detection system for rail surface defects[J].IEEE Transactions on Systems,Man,and Cybernetics,Part C(Applications and Reviews),2012,42(6):1531-1542.
[9] He Zhendong,Wang Yaonan,Yin Feng,et al.Surface Defect Detection for High-speed Rails Using an Inverse P-M Diffusion Model[J].Sensor Review,2016,36(1):86-97.
[10] Li Qingyong,Ren Shengwei.A Real-Time Visual Inspection System for Discrete Surface Defects of Rail Heads[J].IEEE Transactions on Instrumentation & Measurement,2012,61(8):2189-2199.
[11] 邬锋,茅正冲.铁轨表面缺陷图像增强与分割算法[J].计算机仿真,2015,32(10):159-162.Wu Feng,Mao Zhengchong.Enhancement and Segmentation Algorithm for Rail Surface Defects Image[J].Computer Simulation,2015,32(10):159-162.(in Chinese)
[12] 王海涛,甄理,杨春霞,等.基于计算机视觉的铁轨表面缺陷检测系统[J].无损检测,2011,33(11):38- 41.Wang Haitao,Zhen Li,Yang Chunxia,et al.Rail Surface Defects Dection System based on Computer Vision Technology[J].Nondestructive Detection,2011,33(11):38- 41.(in Chinese)
[13] Cand` es E J,Li Xiaodong,Ma Yi,et al.Robust Principal Component Analysis?[J].Journal of the ACM,2011,58(3):1-37.
[14] Kwak N.Principal Component Analysis based on L1-norm Maximization[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2008,30(9):1672-1680.
[15] Markopoulos P P,Karystinos G N,Pados D A.Optimal Algorithms for L1-subspace Signal Processing[J].IEEE Transactions on Signal Processing,2014,62(19):5046-5058.
[16] Ye Jieping.Generalized low rank approximations of matrices[J].Machine Learning,2005,61(1-3):167-191.
[17] Shi Jiarong,Yang Wei,Zheng Xiuyun.Robust Generalized Low Rank Approximations of Matrices[J].PLoS One,2015,10(9):Art.no.e0138028.
[18] Wang Hengyou,Cen Yigang,He Zhihai,et al.Robust Generalized Low-Rank Decomposition of Multi-matrices for Image Recovery[J].IEEE Transactions on Multimedia,2017,9(5):969-983.
[19] Wang Hengyou,Cen Yigang,He Zhihai,et al.Reweighted Low-Rank Matrix Analysis With Structural Smoothness for Image Denoising[J].IEEE Transactions on Image Processing,2018,27(4):1777-1792.
[20] Cen Yigang,Zhao Ruizhen,Cen Lihui,et al.Defect Inspection for TFT-LCD Images based on the Low-rank Matrix Reconstruction[J].Neurocomputing,2015,149:1206-1215.
[21] 王辉,孙洪.鲁棒主成分分析的铝箔表面缺陷检测方法[J].信号处理,2017,33(4):577-582.Wang Hui,Sun Hong.Robust Principal Component Analysis for Aluminum Foil Surface Defects Detection[J].Journal of Signal Processing,2017,33(4):577-582.(in Chinese)
[22] 冯奇,黄建军,张源,等.LCD 缺陷检测系统的图像压缩感知方法[J].信号处理,2018,34(1):72- 80.Feng Qi,Huang Jianjun,Zhang Yuan,et al,An Image Compressive Sensing Method for LCD Mura Detection System[J].Journal of Signal Processing,2018,34(1):72- 80.(in Chinese)
[23] Li Peng,Liang Junli,Shen Xubang,et al,Textile Fabric Defect Detection based on Low-rank Representation[J].Multimedia Tools and Applications,2017(3):1-26.
[24] Cao Junjie,Wang Nannan,Zhang Jie,et al.Detection of Varied Defects in Diverse Fabric Images via Modified RPCA with Noise Term and Defect Prior[J].International Journal of Clothing Science and Technology,2016,28(4):516-529.
[25] Zhang Duo,Gao Guangshuai,Li Chunlei.Fabric Defect Detection Algorithm based on Gabor Filter and Low-rank Decomposition[C]//Eighth International Conference on Digital Image Processing.International Society for Optics and Photonics,2016.
[26] Wright J,Peng Yigang,Ma Yi,et al.Robust Principal Component Analysis:Exact Recovery of Corrupted Low-rank Matrices by Convex Optimization[C]//Advances in Neural Information Processing Systems,2009,3.
[27] Lin Zhouchen,Chen Minming,Ma Yi.The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices[EB/OL].Eprint Arxiv,2010,9.
[2] 戴鹏,王胜春,杜馨瑜,等.基于半监督深度学习的无砟轨道扣件缺陷图像识别方法[J].中国铁道科学,2018,39(4):45-51.Dai Peng,Wang Shengchun,Du Xinyu,et al.Image recognition Method for the Fastener Defect of Ballastless Track based on Semi-Supervised Deep Learning[J].China Railway Science,2018,39(4):45-51.(in Chinese)
[3] 刘孟轲,吴洋,王逊.基于卷积神经网络的轨道表面缺陷检测技术实现[J].现代计算机,2017,10(中):65- 69.Liu Mengke,Wu Yang,Wang Xun.Implementation of Track Surface Defect Inspection Based on Convolutional Neural Network[J].Modern Computer,2017,10(Middle):65- 69.(in Chinese)
[4] 茅正冲,邬锋.基于图像传感器的铁轨表面缺陷视觉检测算法[J].传感器与微系统,2015,34(9):141-144.Mao Zhengchong,Wu Feng.Visual Detection Algorithm for Rail Surface Defects based on Image Sensor[J].Transducer and Microsystem Technologies,2015,34(9):141-144.(in Chinese)
[5] 刘雄祥.基于卷积神经网络的铁轨表面缺陷识别研究[D].西南科技大学硕士论文,2018.Liu Xiongxiang.Research on Rail Surface Defect Recognition based on Convolutional Neural Network[D].Thesis for Master Degree of the Southwest University of Science and Technology,2018.(in Chinese)
[6] 宋亚帆,潘迪夫,韩锟.复杂环境下基于视觉显著性特征的铁轨识别方法[J].铁道科学与工程学报,2018,15(4):57- 65.Song Yafan,Pan Difu,Han Kun.Track Detection Approach in Complex Environment based on Saliency Features[J].Journal of Railway Science and Engineering,2018,15(4):57- 65.(in Chinese)
[7] Gan Jinrui,Li Qingyong,Wang Jianzhu,et al.A Hierarchical Extractor-Based Visual Rail Surface Inspection System[J].IEEE Sensors Journal,2017,17(23):7935-7944.
[8] Li Qingyong,Ren Shengwei.A visual detection system for rail surface defects[J].IEEE Transactions on Systems,Man,and Cybernetics,Part C(Applications and Reviews),2012,42(6):1531-1542.
[9] He Zhendong,Wang Yaonan,Yin Feng,et al.Surface Defect Detection for High-speed Rails Using an Inverse P-M Diffusion Model[J].Sensor Review,2016,36(1):86-97.
[10] Li Qingyong,Ren Shengwei.A Real-Time Visual Inspection System for Discrete Surface Defects of Rail Heads[J].IEEE Transactions on Instrumentation & Measurement,2012,61(8):2189-2199.
[11] 邬锋,茅正冲.铁轨表面缺陷图像增强与分割算法[J].计算机仿真,2015,32(10):159-162.Wu Feng,Mao Zhengchong.Enhancement and Segmentation Algorithm for Rail Surface Defects Image[J].Computer Simulation,2015,32(10):159-162.(in Chinese)
[12] 王海涛,甄理,杨春霞,等.基于计算机视觉的铁轨表面缺陷检测系统[J].无损检测,2011,33(11):38- 41.Wang Haitao,Zhen Li,Yang Chunxia,et al.Rail Surface Defects Dection System based on Computer Vision Technology[J].Nondestructive Detection,2011,33(11):38- 41.(in Chinese)
[13] Cand` es E J,Li Xiaodong,Ma Yi,et al.Robust Principal Component Analysis?[J].Journal of the ACM,2011,58(3):1-37.
[14] Kwak N.Principal Component Analysis based on L1-norm Maximization[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2008,30(9):1672-1680.
[15] Markopoulos P P,Karystinos G N,Pados D A.Optimal Algorithms for L1-subspace Signal Processing[J].IEEE Transactions on Signal Processing,2014,62(19):5046-5058.
[16] Ye Jieping.Generalized low rank approximations of matrices[J].Machine Learning,2005,61(1-3):167-191.
[17] Shi Jiarong,Yang Wei,Zheng Xiuyun.Robust Generalized Low Rank Approximations of Matrices[J].PLoS One,2015,10(9):Art.no.e0138028.
[18] Wang Hengyou,Cen Yigang,He Zhihai,et al.Robust Generalized Low-Rank Decomposition of Multi-matrices for Image Recovery[J].IEEE Transactions on Multimedia,2017,9(5):969-983.
[19] Wang Hengyou,Cen Yigang,He Zhihai,et al.Reweighted Low-Rank Matrix Analysis With Structural Smoothness for Image Denoising[J].IEEE Transactions on Image Processing,2018,27(4):1777-1792.
[20] Cen Yigang,Zhao Ruizhen,Cen Lihui,et al.Defect Inspection for TFT-LCD Images based on the Low-rank Matrix Reconstruction[J].Neurocomputing,2015,149:1206-1215.
[21] 王辉,孙洪.鲁棒主成分分析的铝箔表面缺陷检测方法[J].信号处理,2017,33(4):577-582.Wang Hui,Sun Hong.Robust Principal Component Analysis for Aluminum Foil Surface Defects Detection[J].Journal of Signal Processing,2017,33(4):577-582.(in Chinese)
[22] 冯奇,黄建军,张源,等.LCD 缺陷检测系统的图像压缩感知方法[J].信号处理,2018,34(1):72- 80.Feng Qi,Huang Jianjun,Zhang Yuan,et al,An Image Compressive Sensing Method for LCD Mura Detection System[J].Journal of Signal Processing,2018,34(1):72- 80.(in Chinese)
[23] Li Peng,Liang Junli,Shen Xubang,et al,Textile Fabric Defect Detection based on Low-rank Representation[J].Multimedia Tools and Applications,2017(3):1-26.
[24] Cao Junjie,Wang Nannan,Zhang Jie,et al.Detection of Varied Defects in Diverse Fabric Images via Modified RPCA with Noise Term and Defect Prior[J].International Journal of Clothing Science and Technology,2016,28(4):516-529.
[25] Zhang Duo,Gao Guangshuai,Li Chunlei.Fabric Defect Detection Algorithm based on Gabor Filter and Low-rank Decomposition[C]//Eighth International Conference on Digital Image Processing.International Society for Optics and Photonics,2016.
[26] Wright J,Peng Yigang,Ma Yi,et al.Robust Principal Component Analysis:Exact Recovery of Corrupted Low-rank Matrices by Convex Optimization[C]//Advances in Neural Information Processing Systems,2009,3.
[27] Lin Zhouchen,Chen Minming,Ma Yi.The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices[EB/OL].Eprint Arxiv,2010,9.