基于边缘特征的无砟轨道扣件定位方法

A Locating Method for Ballastless Track Fasteners Based on Edge Feature

摘要:

轨道扣件的精确定位是实现扣件缺陷检测的前提,而常规的基于机器视觉的轨道扣件定位检测方法存在适应性差,且易受光照强度及遮挡的干扰。为了实现轨道扣件的快速精确定位,提出一种基于边缘特征的轨道扣件定位方法。通过在参考图像中选取标准扣件区域以生成匹配模板,利用Canny边缘滤波算子获得边缘点的位置坐标及梯度方向；在此基础上构建搜索模型并采用图像金字塔匹配搜索策略,获得匹配分值的潜在匹配点；并利用匹配阈值设定法,逐层逐次跟踪潜在匹配点,直至图像金字塔最底层,以提高定位速度；基于最小二乘法调整位姿参数,使其达到亚像素级精度。实验表明,该方法具有鲁棒性强、定位速度快且不受光照变化及遮挡的影响,定位精度达到1/15像素,定位成功率大于95%,满足无砟轨道扣件定位需求。

Abstract:

The precise positioning of the track fasteners is the prerequisite for the detection of fastener defects, while the conventional method of locating the track fastener based on machine vision is poor in adaptability and is susceptible to light intensity and shielding.In order to realize the rapid and accurate positioning of the track fastener, a method of locating the rail fastener based on the edge feature is proposed.By using the Canny edge filter operator, the position coordinates and the gradient direction of the edge point are obtained by selecting the standard fastener region in the reference image. On this basis, the search model is constructed and the image pyramid matching search strategy is used to obtain the matching score And then use the matching threshold setting method to track the potential matching points one by one until the bottom of the image pyramid to improve the positioning speed. Based on the least squares method to adjust the pose parameters to achieve sub-pixel-level accuracy.The experimental results show that the method is robust and has good positioning speed and is not affected by illumination and occlusion. The positioning accuracy is 1/15 pixel, and the positioning success rate is more than 95%, which satisfies the positioning requirement of ballastless track fasteners.