基于计算机视觉的钢轨扣件检测算法研究

摘要

针对传统钢轨检测方法不能满足线路检修的需要,提出了一种基于计算机视觉的钢轨扣件检测算法,运用投影法和特定区域像素点扫描统计相结合的方法定位扣件位置,使用灰度特征和HOG特征描述扣件特征向量,并利用Chi开方距离分类器进行特征提取.实验结果表明,该算法具有一定的有效性和可行性.%As the traditional rail detection method can no longer meet the railway maintenance requirements, an detection algorithm of rail fastening based on computer vision is proposed in this paper. The position of the fas-tener can be located by using the projection method and the method of scanning pixels and statistics of specific areas. The characteristics of fasteners are described by way of gray level features and HOG features, and the Chi square distance classifier is adopted to extract features. Results indicate that the algorithm shows certain validity and feasibility.