基于局部结构化特征稀疏编码的手制动机故障检测  

Fault detection of hand brake chain based on local structured characteristic sparse coding

作　　者：刘盛亚 Philip Yamba 邹荣[1] 许桢英[1] 崔世林[2] LIU Shengya;Philip Yamba;ZOU Rong;XU Zhenying;CUI Shilin(School of Mechanical Engineering,Jiangsu University,Zhenjiang 212013,China;School of Electronic and Electric,Nanyang Institute of Science and Technology,Nanyang 473004,China)

机构地区：[1]江苏大学机械工程学院,江苏镇江212013 [2]南阳理工学院电子与电气工程学院,河南南阳473004

出　　处：《铁道科学与工程学报》2018年第9期2415-2422,共8页Journal of Railway Science and Engineering

基　　金：河南省科技攻关项目(172102210414);江苏省博士后科研资助计划项目(1402012B);江苏大学高级专业人才科研启动基金资助项目(14JDG134)

摘　　要：列车机械部件故障检测为铁路安全运营提供保障,而机器视觉技术的发展使得目标故障检测成为检测的主要手段。针对铁路货车手制动机链条丢失故障,由前端视觉图像传感器采集手制动机链条图像,并根据链条具有丰富纹理结构的特点,提出一种新颖的低层次结构化特征;通过稀疏编码构建出中等层次结构化特征,在空间金子塔架构下实现手制动机链条丢失故障的检测。由于结构化特征隐含了图像中的空间结构关系,使得故障检测性能获得极大提高,相应的实验表明,在线性SVM下的故障检测率达到了98%左右,而检测速度达到了9帧/s,具有很好的实时性和很高的检测精度。Fault detection of train mechanical parts provides guarantee for railway safety operation,and the development of machine vision technology makes the target fault detection become the main means of detection.According to the characteristics of the rich texture structure of the chain,this paper presented a novel low-level structured feature,which was based on the characteristics of the low-level structure of the railway truck brake chain.It was composed of the front-end visual image sensor and coding to build a medium-level structured features in the space of the tower under the framework of the realization of hand brake chain fault detection.Because the structural features imply the spatial structure relation in the image,the fault detection performance is greatly improved.The corresponding experiments show that the fault detection rate under the linear SVM is about 98%,and the detection speed reaches 9 frames/sec,with good real-time and high detection accuracy.

关 键 词：视觉故障检测 手制动机链条 局部结构化特征 稀疏编码 空间金字塔 

分 类 号：TH7[机械工程—仪器科学与技术]