基于Faster R-CNN的钢轨表面缺陷识别研究

doi:10.16180/j.cnki.issn1007-7820.2020.09.011

电子科技 ›› 2020, Vol. 33 ›› Issue (9): 63-68.doi: 10.16180/j.cnki.issn1007-7820.2020.09.011

基于Faster R-CNN的钢轨表面缺陷识别研究

苏烨¹,李筠¹,杨海马¹,刘瑾²,江声华³

1. 上海理工大学光电信息与计算机工程学院,上海 200093
2. 上海工程技术大学电子电气工程学院,上海201620
3. 上海瑞纽机械股份有限公司,上海 201314

收稿日期:2019-06-17 出版日期:2020-09-15 发布日期:2020-09-12
作者简介:苏烨(1995-),男,硕士研究生。研究方向:智能检测与控制。|杨海马(1979-),男,博士,副教授。研究方向:光电检测设备开发及信号处理算法应用。
基金资助:
国家自然科学基金(61701296);国家自然科学基金(U1831133);上海市自然科学基金(17ZR1443500);上海航天科技创新基金(SAST2017-062);宝山区科技创新专项基金(17-C-21)

Research on Rail Surface Defect Recognition Based on Faster R-CNN

SU Ye¹,LI Jun¹,YANG Haima¹,LIU Jin²,JIANG Shenghua³

1. School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China
2. School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China
3. Shanghai Ruiniu Machinery Crop,Shanghai 201314,China

Received:2019-06-17 Online:2020-09-15 Published:2020-09-12
Supported by:
National Natural Science Foundation of China(61701296);National Natural Science Foundation of China(U1831133);Shanghai Natural Science Foundation(17ZR1443500);Shanghai Aerospace Science and Technology Innovation Fund(SAST2017-062);Baoshan District Science and Technology Innovation Special Fund(17-C-21)

摘要/Abstract

摘要：

外界因素常会干扰钢轨表面缺陷检测仪器,导致其精度和效率降低。文中研究了一种基于Faster R-CNN网络检测钢轨表面缺陷的方法。该方法将预处理后的图像进行反转,利用Radon变换实现钢轨图像的投影。投影曲线中,利用钢轨长度为定值且灰度值小于图像平均值的特性,完成对钢轨表面区域的提取。然后通过区域建议网络提取候选区域,并与Fast R-CNN网络的区域建议对比分析,完成Faster R-CNN网络对钢轨的表面缺陷检测。试验数据表明,裂缝、疤痕、磨损和划伤4种缺陷的识别精度分别为92.17%、91.85%、93.45%和93.27%,证明使用该方法能够高效而又准确地识别钢轨的表面缺陷。

关键词: 钢轨表面缺陷, 预处理, Radon变换, 灰度值, 区域建议网络, Faster R-CNN网络

Abstract:

External factors usually have effects on the instrument used to detect rail surface defect, resulting in poor accuracy and efficiency of instrument. For this problem, a method for detecting rail surface defects based on the Faster R-CNN network is investigated. The method reverses the preprocessed image and realizes the projection of rail image with Radon transform. In the projection curve, the rail surface area is extracted by using the characteristics that the rail length is fixed and the gray value is less than the average value of the image. Then, the candidate region is extracted through the regional proposal network and compared with the regional recommendation of Fast R-CNN network for the detection of surface defects of rail by the Faster R-CNN networks. According to the test data, the accuracy of crack, scar, abrasion and scratch is 92.17%, 91.85%, 93.45% and 93.27% respectively, which verifies the efficiency and accuracy of the proposed method in identify the surface defects of rail.

Key words: rail surface defect, preprocessing, radon transform, grey value, regional proposal network, Faster R-CNN networks

中图分类号:

TP278

苏烨,李筠,杨海马,刘瑾,江声华. 基于Faster R-CNN的钢轨表面缺陷识别研究[J]. 电子科技, 2020, 33(9): 63-68.

SU Ye,LI Jun,YANG Haima,LIU Jin,JIANG Shenghua. Research on Rail Surface Defect Recognition Based on Faster R-CNN[J]. Electronic Science and Technology, 2020, 33(9): 63-68.

图/表 10

图1

图2

图3

图4

图5

图6

表1

图7

表2

表3

参考文献 16

[1]	孟繁顺. 浅谈铁路钢轨断裂的原因及预防措施[J]. 铁道建筑技术, 2011(10):58-62.
	Meng Fanshun. Thereasons and preventive measures of railway rail fracture are briefly discussed[J]. Railway Construction Technology, 2011(10):58-62.
[2]	Utrata D, Clark R. Groundwork for rail flaw detection using ultras onic phased array inspection[J]. Review of Progress in Quatitative Nondestructive Evaluation, 2003,657(2):799-805.
[3]	Papaelias M P, Lugg M C, Roberts C, et al. High-speed inspection of rails using ACFM techniques[J]. NDT&E International, 2009,42(4):328-335. doi: 10.1016/j.ndteint.2008.12.008
[4]	王时丽, 刘桂华. 基于机器视觉的钢轨表面缺陷三维检测[J]. 微型机与应用, 2015,34(19):10-13.
	Wang Shili, Liu Guihua. 3D defect detection of rail surface based on machine vision[J]. Microcomputer and Applications, 2015,34(19):10-13.
[5]	汪路明, 金合丽, 周煊勇, 等. 基于机器视觉和脉冲涡流复合的在线钢轨表面缺陷检测系统设计[J]. 浙江树人大学学报(自然科学版), 2018(1):7-11.
	Wang Luming, Jin Heli, Zhou Xuanyong, et al. Design of on-line rail surface defect detection system based on machine vision and pulsed eddy current[J]. Journal of Zhejiang Shuren University(Natural Science Edition), 2018(1):7-11.
[6]	王智. 基于BP网络的钢轨缺陷信号实验分析[J]. 电子制作, 2013(10):31-32.
	Wang Zhi. Experimental analysis of rail defect signal based on BP network[J]. Electronic Manufacture, 2013(10):31-32.
[7]	冉建民. 基于图像和CNN模型的钢轨表面缺陷识别研究[D]. 兰州:兰州交通大学, 2018.
	Ran Jianmin. Research on rail surface defect recognition based on image and CNN model[D]. Lanzhou:Lanzhou Jiaotong University, 2018.
[8]	任俊箫. 基于图像处理的铁轨表面缺陷检测设计[D]. 绵阳:西南科技大学, 2017.
	Ren Junxiao. Design of rail surface defect detection based on image processing[D]. Mianyang: Southwest University of Science and Technology, 2017.
[9]	林萌, 李翠华, 黄剑航. 基于Radon变换的运动模糊图像参数估计[J]. 计算机技术与发展, 2008,18(1):33-36.
	Lin Meng, Li Cuihua, Huang Jianhang. Parameter estimation of motion blur image based on Radon transform[J]. Computer Technology and Development, 2008,18(1):33-36.
[10]	Gael R, Jean-Baptiste B, Gerard B, et al. Reflective imaging solved by the Radontranform[J]. IEEE Geoscience & Remote Sensing Letters, 2016,13(7):936-938.
[11]	Girshick R, Donahue J, Darrell T, et al. Region-based convolutional networks for accurate object detection and segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,38(1):1-8. doi: 10.1109/tpami.2015.2439257 pmid: 27030844
[12]	Girshick R. Fast R-CNN[C]. Boston:IEEE International Conference on Computer Vision(ICCV), 2015.
[13]	刘雄祥. 基于卷积神经网络的铁轨表面缺陷识别研究[D]. 绵阳:西南科技大学, 2018.
	Liu Xiongxiang. Research on track surface defect recognition based on convolutional neural network[D]. Mianyang:Southwest University of Science and Technology, 2018.
[14]	Shaoqing R, Kaiming H, Ross G, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017,39(3):1137-1149. doi: 10.1109/TPAMI.2016.2577031
[15]	张汇, 杜煜, 宁淑荣, 等. 基于Faster RCNN的行人检测方法[J]. 传感器与微系统, 2019,38(2):147-149,153.
	Zhang Hui, Du Li, Ning Shurong, et al. Pedestrian detection method based on faster RCNN[J]. Sensors and Microsystems, 2019,38(2):147-149,153.
[16]	刘蕴辉, 刘铁, 王权良, 等. 基于图像处理的铁轨表面缺陷检测算法[J]. 计算机工程, 2007(11):236-238.
	Liu Yunhui, Liu Tie, Wang Quanliang, et al. Image processing based on the track surface defect detection algorithm[J]. Computer Engineering, 2007(11):236-238.

钢轨表面缺陷		平均检测时间/s
疤痕	91.43	0.32
裂缝	91.09
划伤	92.78
磨损	92.46

检测方法	四种缺陷的平均识别准确率/%
检测方法	疤痕	裂缝	划伤	磨损
BP神经网络	87.92	85.26	88.74	83.42
文献[16]	84.35	79.33	82.46	83.75

基于Faster R-CNN的钢轨表面缺陷识别研究

Research on Rail Surface Defect Recognition Based on Faster R-CNN

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 16

相关文章 14

Metrics

本文评价

推荐阅读 10

[1]	司明明,陈玮,胡春燕,尹钟. 融合Resnet50和U-Net的眼底彩色血管图像分割[J]. 电子科技, 2021, 34(8): 19-24.
[2]	晏子杰,王京梅,陈卓,刘宇. 软件可编程的FPGA网络测量引擎技术实现[J]. 电子科技, 2021, 34(2): 27-32.
[3]	葛靖,刘子龙. 基于CNN和LSTM的睡眠呼吸暂停检测算法[J]. 电子科技, 2021, 34(2): 21-26.
[4]	朱健安,魏云冰,朱鹏杰,蒋成成,朱成名. 基于优化灰色傅里叶残差修正的中长期负荷预测[J]. 电子科技, 2021, 34(12): 49-55.
[5]	孙腾飞,蒋红海. 基于模式分类的喷嘴料液雾滴速度测量研究[J]. 电子科技, 2019, 32(8): 33-37.
[6]	毕灶荣，童东兵，陈巧玉. 基于快速MFCC计算的说话人识别系统的设计[J]. , 2018, 31(4): 25-.
[7]	王松会,杨姗姗,李云娟. 中原地区古代陶器数据预处理研究[J]. , 2015, 28(9): 27-.
[8]	刘桥,李晗,何凌杰. 基于光流算法的医学柔式内窥镜系统[J]. , 2015, 28(4): 61-.
[9]	桂腾叶,陈硕,隗立志,崔志军,周晓琳. 基于ARIMA的传染病流行趋势预测及防治对策[J]. , 2015, 28(12): 48-.
[10]	李晓峰,张立军,胡永春. 面向H.264编解码系统的图像预处理技术[J]. , 2015, 28(10): 111-.
[11]	邵晓鹏,靳振华,王阳. 去除红外图像条带噪声改进算法研究[J]. , 2013, 26(10): 83-.
[12]	李佰鹤,魏计林,邱选兵,关利乐. FBG传感器微弱信号预处理[J]. , 2013, 26(1): 45-.
[13]	冯永胜, 郭立新, 刘忠玉. 基于射线跟踪算法数字地图自动提取与预处理[J]. , 2012, 25(10): 4-.
[14]	苏军. 基于频谱分析的运动模糊图像的参数鉴别[J]. , 2011, 24(7): 77-.