中国铁道科学2011，Vol.32Issue(1)：25-29,5.
鲁棒实时钢轨表面擦伤检测算法研究
Research on Robust Fast Algorithm of Rail Surface Defect Detection
摘要
Abstract
By applying the advanced technologies of image processing and pattern recognition, analyzing the digital rail image acquired by high speed linear-scanned camera, a robust fast algorithm of rail surface defect detection was proposed.Firstly, the exact rail region was extracted by analyzing the vertical projection of input images.Then the gray contrast image was transformed from the gray rail image utilizing the model of contrast perception of human being, and the suspected defect was segmented through binary processing based on the maximum entropy principle.At last, the defects were determined with the help of some prior knowledge.The experiment results demonstrate that the algorithm has excellent detection performance.Its average accuracy rate is 90.7％, and the average missing rate is 3.95％.Furthermore, the algorithm has high detection speed, and its mean detection time is no more than 40 ms.关键词
钢轨表面擦伤/检测/对比度感知/最大熵原理/算法分类
交通工程引用本文复制引用
任盛伟,李清勇,许贵阳,韩强,罗四维,冯其波..鲁棒实时钢轨表面擦伤检测算法研究[J].中国铁道科学,2011,32(1):25-29,5.基金项目
国家"八六三"计划项目(2009AA110301) （2009AA110301）
