中文摘要：

针对高速铁路接触网支撑装置旋转双耳部件销钉的松脱与脱落问题,提出一种基于图像不变性目标定位及灰度分布规律特征的销钉不良状态检测方法。通过分析现场接触网支撑及悬挂装置图像,利用SIFT（Scale Invariant Feature transform）算法和改进的RANSAC（Random Sample Consensus）算法实现双耳部件的定位;采用Hough变换实现目标图像中双耳套筒倾角的提取,并将其旋转至水平方向,进而实现旋转双耳部分的分割;累加目标图像的竖直方向像素灰度值,确定销钉受力部分和两端非受力部分长度;归纳销钉正常工作及故障时这些长度间相关比值的范围,从而判断销钉的工作状态。实验表明,该方法能够较准确地实现销钉不良状态的检测。

英文摘要：

In response to the loosening and falling off of swivel clevis pins of the support devices of overhead contact system（OCS）,this paper presented a method to detect the defective conditions of the pins based on target positioning of the invariance of the image and grayscale distribution characteristics.Through the analysis of the obtained OCS support and suspension images on site,the SIFT（Scale Invariant Feature Transform）algorithm and the improved RANSAC（Random Sample Consensus）algorithm were adopted to realize the positioning of swivel clevis.Besides,the Hough transform was adopted to obtain the angle of the clevis end holder in the target image,to rotate the clevis end holder to the horizontal direction,and to achieve the separation of the pins part.The pixel gray values of the vertical direction of target image were cumulated,to determine the lengths of the stress part of the pins and non-stress parts on both ends of the pins.With the scope of related ratios of the lengths concluded on the normal and defective conditions of the pins,the conditions of the pins can be determined.The experiments showed that the method can realize accurate detection of defective conditions of pins.