基于偏振图像融合的接触网U型抱箍螺母松脱检测研究  

Detection of loose nuts for U-shaped hoops of catenary by using polarized image fusion

作　　者：刘仕兵 全丰 喻星 LIU Shibing;QUAN Feng;YU Xing(School of Electrical and Automation Engineering,East China Jiaotong University,Nanchang 330013,China)

机构地区：[1]华东交通大学电气与自动化工程学院,江西南昌330013

出　　处：《铁道科学与工程学报》2021年第12期3357-3366,共10页Journal of Railway Science and Engineering

基　　金：江西省科技厅重点研发项目(20192BBH80005)。

摘　　要：为了解决高铁接触网腕臂上U型抱箍螺母松脱问题,提出一种偏振成像与改进余弦算法结合的故障诊断方法。引入对比量互信息(MI)融合2个偏振参数,研究偏振技术的应用特性。提升小波融合规则中,在其低频系数中引入PCNN(Pulse Coupled Neural Network)耦合网络进行点火选择,高频系数以空间频率MSF特征量为比较,重构系数后融合偏振特征图与强度图像。将新图像变换为LBP旋转不变模式并将特征值映射到cosine向量中,通过手工设计提取特征对螺母进行状态检测。研究结果表明:该方法下偏振融合图像轮廓鲜明,纹理细节及各图像评价指标均有提高。通过对故障样本进行试验检测其准确率达到90.9%,对后期接触网系统检测技术具有参考意义。In order to solve the looseness problem of the U-shaped hoops on high-speed railway catenary, the combination of polarized images and improved cosine method was proposed for looseness detection. The characteristics of polarized technology and the inclusion of contrastive mutual information for fusing two polarized parameters were first studied. Then improved wavelet fusion rules for selecting Low-frequency coefficients were introduced by the Pulse Coupled Neural Network(PCNN), and high-frequency coefficients were compared with MSF. Finally, fused images were transformed into LBP rotation invariant mode whose eigenvalues were mapped to the cosine vector, and the states of hoops and nuts were detected through extracting features by manual design. The research results indicate that polarized fusion images have a clear outline,enhanced texture details, and improved image evaluation indicators by this method. The accuracy of fault detection on the samples reaches 90.9%. It has important reference value for the evaluation and detection of catenary system problems.

关 键 词：接触网 U型抱箍 偏振 小波融合 余弦相似 

分 类 号：U225[交通运输工程—道路与铁道工程] TP391[自动化与计算机技术—计算机应用技术]