To address the issue that jagged edges of the wheel-rail contact surface lead to low accuracy of segmentation algorithm for grayscale rail images, this paper proposes a segmentation method for wheel-rail contact surface in RGB rail images based on dynamic snake convolution. A rail image acquisition module based on a color linear array camera and a white laser light source is designed and implemented. By embedding dynamic snake convolution, this research enhances the extraction of jagged irregular features and improves the DeepLabv3+ segmentation network, thereby completing integrated segmentation of wheel-rail contact surface and rail surface based on RGB image, as well as classification and detection of the wheel-rail contact surface. The experimental results show that the average Intersection over Union (IoU) for RGB rail images excluding turnout section segmentation is 93.5%, the average pixel accuracy of the category is 96.39%, and the pixel accuracy is 98.85%. For RGB images containing the turnout section segmentationt, the average IoU, the average pixel accuracy of the category and the pixel accuracy of the are 91.87%, 96.04%, and 98.60%, respectively. RGB images can better represent the true state of the wheel-rail contact surface. The segmentation network improved by adding dynamic snake convolution can enhance the accurate extraction of the rail-wheel rail contact surface area, the average IoU is improved by 2.25% compared to existing methods.
HUErgen. Causes and Prevention of Rail Abrasion [J]. Railway Engineering, 2000, 40 (1): 32-33. in Chinese
[3]
罗林,张格明,吴旺青,等.轮轨系统轨道平顺状态的控制[M].北京:中国铁道出版社,2006.
[4]
LUOLin, ZHANGGeming, WUWangqing, et al. Control of Track Regularity in Wheel-Rail Systems [M]. Beijing: China Railway Publishing House, 2006. in Chinese
LIUFengshou, LIChuang, TIANChanghai. Study on Early Damage of Rail in High Speed Railway in China [J]. Railway Engineering, 2018, 58 (1): 138-140. in Chinese
LIChuang, ZHANGYinhua, TIANChanghai, et al. Study on Rail Service Status and Disease Treatment of Rail for High Speed Railway [J]. Railway Engineering, 2020, 60 (8): 126-129, 142. in Chinese
[9]
赵康云.高速铁路钢轨常见表面伤损及其对策分析[J].中国高新科技,2020(17):83-84.
[10]
ZHAOKangyun. Common Surface Damage of High-Speed Railway Rails and Its Countermeasures [J]. China High and New Technology, 2020 (17): 83-84. in Chinese
XUGuiyang, SHITianyun, RENShengwei, et al. Development of the On-Board Track Inspection System Based on Computer Vision [J]. China Railway Science, 2013, 34 (1): 139-144. in Chinese
[13]
左玉良.从钢轨光带异常看高速线路轨道平顺性[J].铁道标准设计,2009,53(4):13-16.
[14]
ZUOYuliang. Study on Smoothness of High Speed Railway Track from Unusualness of Steel Track Light Ribbon [J]. Railway Standard Design, 2009, 53 (4): 13-16. in Chinese
XIEMinjie, FengkunLÜ, YUANXiaocui. Design of Rail Surface Defect Detection System Based on Machine Vision [J]. Journal of Nanchang Institute of Technology, 2020, 39 (1): 74-79. in Chinese
LIZhonghai, BAIQiuyang, WANGFuming, et al. Real-Time Detection System of Rail Surface Defects Based on Semantic Segmentation [J]. Computer Engineering and Applications, 2021, 57 (12): 248-256. in Chinese
WULushen, ZHANGCong, WANChao, et al. Study on Optical Model for Rail Surface Detection Based on Machine Vision [J]. Railway Standard Design, 2017, 61 (5): 50-57. in Chinese
TANSong, LIWeiwei, HANQiang. Research and Manufacture of On-Board Detection System for Abnormal Rail Light Band Based on Computer Vision [J]. Railway Engineering, 2016, 56 (2): 128-131. in Chinese
MAOQingzhou, KOUDonghua, DAIYongbo, et al. Research on Key Technologies of Intelligent Detection of Railway Rail Appearance and Geometric State [J]. Railway Rolling Stock, 2021, 59 (6): 64-72. in Chinese
LIWeiyi, HANQiang, WANGDengyang, et al. Inspection System for Light Band of High Speed Rails [J]. China Railway, 2018 (2): 100-104. in Chinese
[27]
张博,刘秀波.基于机器视觉的钢轨光带检测方法[J].铁道建筑,2022,62(10):34-44.
[28]
ZHANGBo, LIUXiubo. Detection Method of Running Band of Rail Based on Machine Vision [J]. Railway Engineering, 2022, 62 (10): 34-44. in Chinese
[29]
BOJARCZAKP, NOWAKOWSKIW. Application of Deep Learning Networks to Segmentation of Surface of Railway Tracks [J]. Sensors, 2021, 21 (12): 4065-4065.
[30]
LONGJ, SHELHAMERE, DARRELLT. Fully Convolutional Networks for Semantic Segmentation [C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA, USA. New York: IEEE, 2015: 3431-3440.
[31]
WANGQ H, GAOT C, HEQ, et al. Severe Rail Wear Detection with Rail Running Band Images [J]. Computer-Aided Civil and Infrastructure Engineering, 2023, 38 (9): 1162-1180.
[32]
GUOF, QIANY, YUH Y. Automatic Rail Surface Defect Inspection Using the Pixelwise Semantic Segmentation Model [J]. IEEE Sensors Journal, 2023, 23 (13): 15010-15018.
[33]
CHENL C, ZHUY K, PAPANDREOUG, et al.Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation [M]// European Conference on Computer Vision - ECCV 2018. Cham: Springer, 2018: 833-851.
[34]
VASWANIA, SHAZEERN, PARMARN, et al. Attention is All You Need [C]// 31st Conference on Neural Information Processing Systems (NIPS 2017). California, USA. New York: Curran Associates Inc., 2017, 6000-6010.
ZHANGXinfeng, BIANHaonan, ZHANGBo, et al. Rail Light Band Detection Algorithm Based on Deep Learning [J]. Computer Science, 2023, 50 (): 301-306. in Chinese
RENCheng. An Improved Douglas-Peucker Algorithm Based on Radial Constraint and Point Optimization [J]. Surveying and Mapping of Geology and Mineral Resources, 2017, 33 (4): 5-7. in Chinese
[39]
QIY L, HEY T, QIX M, et al. Dynamic Snake Convolution Based on Topological Geometric Constraints for Tubular Structure Segmentation [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Paris, France. Paris: IEEE, 2023: 6070-6079.
CHENGYu, DUXinyu, GUZichen, et al. High-Speed Real-Time Acquisition and Processing System of Track Inspection Images Based on FPGA and DSP [J]. China Railway Science, 2021, 42 (1): 32-42. in Chinese
[42]
HAOY Y, LIUY, CHENY Z, et al. EISeg: an Efficient Interactive Segmentation Tool Based on PaddlePaddle [EB/OL]. 2022: 2210.08788.