关键词: 钢轨, 光带异常, 检测

Abstract: 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 reach 220 km/h.

Key words: rail, abnormal light band, inspection