特殊环境下接触网精测装置障碍物自动识别算法  

Automatic obstacle recognition algorithm for precise measurement equipment of overhead lines in special environments

作　　者：樊衍 张磊 吴文平 FAN Yan;ZHANG Lei;WU Wen-ping(Guoneng Shuohuang Railway Development Co.,Ltd.,Yuanping 034100,Shanxi Province,China)

机构地区：[1]国能朔黄铁路发展有限责任公司,山西原平034100

出　　处：《信息技术》2025年第2期156-161,167,共7页Information Technology

摘　　要：在接触网精测过程中,由于精测装置对特殊环境干扰因素较为敏感,使得障碍物识别结果AUC值较低。因此,提出特殊环境下接触网精测装置障碍物自动识别算法。定义坐标转换公式,标定接触网精测装置。采集接触网图像并进行滤波和增强,获取图像边缘信息。结合径向运动补偿原理和像素值重分配算法,分析预处理后的图像,找到疑似障碍物目标。融合DeblurGAN-v2去模糊网络和YOLOv5s网络,构建障碍物自动识别模型,输出最终障碍物识别结果。实验结果表明:所提算法的障碍物自动识别结果F1值达到了0.82,基本满足了接触网的障碍物检测精度要求。In the process of precise measurement of the overhead lines,due to the sensitivity of the precise measurement device to special environmental interference factors,the AUC value of obstacle identification results is relatively low.Therefore,an automatic obstacle recognition algorithm for the precise measurement device of the overhead lines in special environments is proposed.The coordinate conversion formulas is defined and the precise measurement device of the overhead lines is calibrated.The overhead lines images are collected,which are filtered and enhanced to obtain image edge information.Combining the principle of radial motion compensation and pixel value reassignment algorithm,the preprocessed image are analyzed and the suspected obstacle targets are found.The DeblurGAN-v2 deblurring network and YOLOv5s network are integrated to construct an automatic obstacle recognition model,and output the final obstacle recognition results.The experiment results show that the F1 value of the proposed algorithm for automatic obstacle recognition reaches 0.82,which basically meets the requirements of precise measurement of the contact network.

关 键 词：特殊环境 接触网 精测装置 障碍物 自动识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]