借助弱纹理匹配的TEDS车底故障区域定位算法  

TEDS underbody fault location algorithm in virtue of weak texture matching

作　　者：黄粤豫 周航[1] 陈业泓 陆鑫 余佳 韩睿宇 HUANG Yueyu;ZHOU Hang;CHEN Yehong;LU Xin;YU Jia;HAN Ruiyu(School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学电子信息工程学院,北京100044

出　　处：《智能系统学报》2024年第3期670-678,共9页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金面上项目(61872027);北京交大科研项目(W21L00390);中建电子智能交通研究生联合培养基地建设项目(275210529245).

摘　　要：针对当前动车组运行故障动态图像检测系统(trouble of moving EMU detection system,TEDS)故障识别准确率低的问题,本文提出一种借助弱纹理匹配的动车底部潜在故障区域定位方法。首先,采用拓扑交叉数检测大量弱纹理区域特征点;然后,以特征点为中心的环形区域内各像素点的拓扑交叉数值筛选特征点,构建相应特征向量进行弱纹理特征匹配;最后,对配准后的图像进行比对定位潜在故障区域。实验结果表明,该算法保证了匹配精度,能检测出大部分潜在故障区域,弱纹理区域的特征匹配准确率超过80%且所有图像对均存在特征匹配对,为以后的精准故障分类提供了有利条件。Fault recognition accuracy in the trouble of moving EMU detection system(TEDS)is suboptimal.Thus,a method for locating potential fault areas at the bottom of EMU by means of weak texture matching is proposed.First,a large number of feature points in the weak texture areas are detected using the topological crossover number.Second,feature points are selected on the basis of the topological crossover number of each pixel within a specified ring area centered on the feature points.This approach aims to construct corresponding feature vectors to achieve the matching of weak texture features.Finally,the registered image is compared to locate the potential fault areas.Experimental results show that the algorithm ensures matching accuracy;it can also detect most potential fault areas.The matching accuracy of weak texture areas is more than 80%,with feature matching pairs in all image pairs.Such a result provides favorable conditions for accurate fault classification in the future.

关 键 词：图像配准 特征匹配 弱纹理特征 潜在故障 区域定位 拓扑交叉数 均值漂移 特征筛选 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]