潘得罗扣件的弹条状态识别算法  

Multi-feature Fusion of Pandrol Rapid Elastic Strip Fastener State Recognition Algorithm

作　　者：朱澎程 苏圣超 陈兴杰[1] 李立明[1] ZHU Pengcheng;SU Shengchao;CHEN Xingjie;LI Liming(School of Urban Railway Transportation,Shanghai University of Engineering and Technology,201620,Shanghai,China)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620

出　　处：《城市轨道交通研究》2021年第6期46-51,共6页Urban Mass Transit

基　　金：上海市科委地方能力建设项目(18030501300);国家自然科学基金面上项目(51975347)。

摘　　要：为实现潘得罗扣件的自动化巡检,提出一种融合弹条纹理特征与形状特征的潘得罗扣件的弹条状态识别算法。该算法首先根据人工巡检的经验划分检测区域,然后增强弹条的纹理与轮廓,接着分别提取弹条的纹理特征与形状特征,组成复特征向量后将其放入SVM(支持向量机)分类器中建立训练模型,最后使用SVM分类器识别钢轨扣件的弹条状态。试验结果表明,最终检测效果为:漏检率7.1%,误检率2.9%,正确率92.9%,均达到钢轨扣件的检测要求。To realize automatic inspection of Pandrol fastener,a Pandrol fastener spring state detection algorithm that integrates spring texture and shape features is proposed.The algorithm first divides the detection area according to the experience of manual inspection.Then the texture and contour of the fastened spring are enhanced.Then the texture features and shape features of the spring are extracted separately,and the complex feature vector is composed and put into the SVM classifier to establish the training model.Finally,the SVM classifier is used to identify the status of the track fasteners.The experimental results show that the missed detection rate is 7.1%,the false detection rate is 2.9%,and the correct rate is 92.9%,which meets the detection requirements of track fasteners.

关 键 词：轨道交通 潘得罗扣件 状态识别 

分 类 号：U213.531[交通运输工程—道路与铁道工程]