砂轮地貌双目图像特征点提取和匹配研究

Grinding wheel topography’s feature extraction and matching based on binocular image

特征提取和匹配是双目视觉中的关键点和难点。对SIFT和Harris两种特征点提取算法应用于双目砂轮地貌图像的角点检测进行了研究,通过对SIFT特征采用欧式距离进行匹配,对Harris角点采用零均值归一化互相关算法(ZNCC)进行模板匹配,采用RANSAC算法剔除误匹配特征点对。实验结果表明,SIFT算法在双目砂轮地貌图像上应用较Harris算法效果理想。

Feature extraction and matching is a binocular vision difficulties and key points. In this paper, SIFT and Harris feature point extraction algorithm is applied to two kinds of grinding wheel topography binocular image corner detection has been studied, through the SIFT features are matched using Euclidean distance, on the Harris corner with Zero-mean Normalized Cross-Correlation algorithm （ZNCC）for template matching algorithm using RANSAC feature points excluding the mismatching pair. Experimental results show that, SIFT algorithm in the application of binocular image grinding wheel topography than Harris algorithm is effective.