基于SIFT特征匹配的电力设备外观异常检测方法

基于电力设备巡检机器人平台，提出了一种电力设备外观异常检测方法。该方法使用SIFT特征点提取算法，进行特征匹配，使用基于RANSAC的方法求解单应矩阵进行配准。在经过配准后，对像素的差值使用Mean Shift分割算法提取异常区域。实验证明该方法对于光照有较高的鲁棒性，且受匹配误差和拍摄角度偏差、位置偏差影响较小，能够有效的将变化区域提取。

Research on Abnormal Appearance Detection Approach of Electric Power Equipment

Based on the electric power equipment automatic inspection robot, this paper presents an image processing approach for abnormal appearance of electric power equipment. Using camera images, we first automatically to find the robust feature points to match though SIFT method. Based on the basic match, we use RANSAC algorithms to find the Homography of the test image and the reference image. After the registration, we get the accurate location of the abnormal areas in the test image by Mean Shift segment method. The experimental results show that this method is efficient, and very robust to the angle deviation and position deviation, even the matching error.