基于改进Canny算法的车轮踏面边缘检测  

Wheel Tread Edge Detection Based on Improved Canny Algorithm

作　　者：孙宝民 王冬 林含潇 杨钰鑫 SUN Baomin;WANG Dong;LIN Hanxiao

机构地区：[1]国能铁路装备有限责任公司包头车辆维修分公司生产技术部,内蒙古包头014060 [2]国能铁路装备有限责任公司包头车辆维修分公司万水泉北场运用车间,内蒙古包头014060 [3]西南交通大学轨道交通运载系统全国重点实验室,成都610031

出　　处：《科技创新与应用》2024年第31期17-22,29,共7页Technology Innovation and Application

基　　金：国能铁路装备有限责任公司包头车辆维修分公司研究项目(TZKY-21-45)。

摘　　要：近年来,列车检修逐渐由人检转换为机检,针对在踏面损伤检测图像处理过程中,由于环境光照、背景区域复杂等因素影响,导致图像踏面目标区域边缘检测效果差、细节信息丢失等问题,提出一种基于引导滤波改进Canny算子的边缘检测算法。首先,利用中值滤波和引导滤波组成的混合滤波进行图像去噪处理,在滤除噪声的同时具有良好的保边特性;接着,扩展梯度模板的计算方向来获得梯度幅值;最后,利用边缘信息改进全局阈值的方法自适应获得高低阈值。基于引导滤波改进Canny算子的边缘检测算法能够有效滤除图像噪声,同时检测出更多的踏面损伤边缘信息。In recent years,train maintenance has gradually shifted from human inspection to machine inspection.In view of the problem of poor edge detection in the image processing process of tread damage detection,due to factors such as environmental lighting and complex background areas,the edge detection effect of the image tread target area is poor and the detail information is lost,an edge detection algorithm based on guided filtering to improve the canny operator is proposed.First,a hybrid filter composed of median filter and guided filter is used to denoise the image,which filters out noise and has good edge preservation characteristics;then,the calculation direction of the gradient template is expanded to obtain the gradient amplitude;finally,edge information is used to improve the global threshold method adaptively obtains high and low thresholds.The edge detection algorithm based on guided filtering and improved canny operator can effectively filter out image noise and detect more edge information of tread damage.

关 键 词：列车 车轮踏面 损伤 边缘检测 引导滤波 

分 类 号：U292.9[交通运输工程—交通运输规划与管理]