无砟轨道扣件快速匹配定位方法研究

The Fast Matching and Localization Method for Ballastless Track Fastener

摘要:

针对现有扣件定位方法速度慢、定位不准确的问题，结合无砟轨道的特点，提出一种无砟轨道扣件快速匹配定位算法。该算法针对传统匹配算法运算量大、匹配结果不准确的缺点，采用梯度法求取扣件图像的方向场，并根据钢轨区域方向场的各向一致性，锁定扣件感兴趣区域，实现图像裁剪，减少模板搜索空间；模板匹配则通过计算模板和图像方向场对应点之间的绝对值距离实现扣件匹配定位，并利用方向场采样、基于统计方法确定搜索起点位置、随机抽样一致性等方法提高匹配速度和精度。实验结果表明，该算法匹配速度快、鲁棒性强，能够满足扣件检测识别中对扣件匹配的实时性和准确性的要求，与传统的模板匹配定位方法相比，速度提高了30%左右，且对于不同光线条件下的匹配成功率均大于95%。

Abstract:

Since the existing location algorithm of faster is slow and the result is inaccurate, combining with the characteristics of Ballastless Track Fastener, this paper proposed a fast matching and localization method for ballastless track fastener. In view of the problem of traditional fastener matching method, which is complex and inaccurate, the gradient directional field was used. Moreover, according to the consistency of consistency, the target area of faster can be locked, which may reduce the space of search by cropping the image. Template matching calculate the absolute distance of corresponding points in the directional field between the template and the image, which combining the directional field sampling, the starting point of search based on statistical method, and the method of adjusting the position and attitude parameters by the least square method to improve the matching speed and accuracy of results. The results of experiment show that the method greatly improves matching speed and has the characteristics of robustness, which can meet the demands of real-time and accuracy of matching and localization for ballastless track fastener. Compared with the traditional template matching method, the percent of speed improvements using this way is 30% and the percent of matching success under different lighting conditions is greater than 95%.