铁路钢轨光带异常检测系统

Inspection system for rail abnormal light band

应用图像识别技术,结合钢轨光带结构特征,提出线性时间复杂度的光带异常检测算法。首先使用图像垂直方向加权投影分析的方法,定位钢轨区域。然后基于"由粗到细"的分析策略,综合利用光带的灰度特征、局部梯度特征,逐行扫描图像,精确定位光带边界。最后根据经验知识判定光带异常类型:光带整体宽度异常、光带偏心、光带局部异常。实验验证了算法的有效性和高效性,每小时可以分析220 km数据。

A detection algorithm of abnormal rail light band for rail inspection based on linear time complexity is proposed,with the analysis of the feature of abnormal rail light band structure using the image recognition technology.Firstly,the exact rail region is extracted by analyzing the weighted vertical projection of input images.Then the border of the rail light band is accurately located based on the analytic strategy of coarse-to-fine,with the feature of gray and local gradient of the rail light band.The inspection results,including the overall abnormal of the width,the center line deviation and local anomalous width,are determined based on empirical knowledge.The experiment results demonstrate the effectiveness and efficiency of the algorithm,and the inspection speed of detection can reach220km/h.