中文摘要：

针对货运列车运行故障动态检测系统中闸瓦缺失故障的检测，提出一种基于弧度步长的图像分级匹配方法。截取正常闸瓦图像作为模板，经预处理后计算其轮廓不变矩；在待检图中提取车轮外圆，并于该圆心处新建直角坐标系，取第四象限圆弧，以选定弧度为步长取圆弧上点作为一级遍历限定路径，遍历计算待检图模板窗口的轮廓不变矩，采用归一化相关匹配法对相似性进行度量，然后由一级遍历最大相似值及其坐标计算二级遍历动态区域进行小范围遍历，最终确定闸瓦区域。对比实验结果表明，该算法能有效避免待检图复杂背景对匹配的干扰，保证定位准确的同时极大地提高了检测效率，满足闸瓦缺失故障检测的实时性要求。

英文摘要：

For the missing fault of brake shoe in trouble of moving freight ear detection system, a staged matching algorithm based on radian step is proposed. A regional image which contains normal brake shoe is intercepted as the template, then the contour moment invariants are calculated after template is preprocessed. In the image to be inspected ,the excircle of wheel was extracted. The center point of the circle was set as another origin, where a new coordinates was established. Circular arc in the forth quadrant is chosen, from which the points regarded as determinative ergodic path are calculated by selected radian step. During the first stage of ergodic process, corresponding matching similarity of each ergodic points measured by normalized correlation algorithm is calculated between template and template window in the image to be inspected. Then the second stage of ergodic process＇ region was calculated according to the maximum similarity and its coordinate. Finally the region of brake shoe is located. The experiment results has shown that the algorithm can avoid the interference of complex background effectively, make inspection on brake shoe more accurate and much quicker, and can satisfy the instantaneity of the inspection on brake shoe.