工件圆弧边缘特征点检测与匹配方法

Feature point detecting and matching method of workpiece circular edge

在利用双目视觉测量工件圆弧半径的过程中,圆弧边缘特征点的检测与匹配是至关重要的一环。针对工件圆弧边缘特征点的检测,通过改进的随机抽样一致性(RANSAC)算法拟合椭圆,在求得内点中筛选出边缘特征点;针对工件圆弧边缘特征点的匹配,本文通过极线约束使得两幅图像对应的极线平行从而满足扫描线特性,在此基础上,以确定的椭圆中心为匹配参考点,对圆弧边缘特征点进行逐行匹配。在得出匹配点对后,利用改进的RANSAC算法排除误匹配对。通过实验验证提出方法的可靠性与准确性。

In the process of measuring arc radius of workpiece by using binocular vision, detecting and matching of feature points of arc edge are very important. In order to detect the feature points of the workpiece arc, by improved random sample consensus （RANSAC） algorithm, fit the ellipse, and screen edge feature points in the inner point ; in order to match the feature points of workpiece arc, make epipolar lines corresponding to two images parallel by epipolar constraint, so that the scanning line characteristics can be satisfied. On this basis, the ellipse center is chosen as the matching reference point, and arc edge feature points are matched by line. After getting the matching points,the improved RANSAC algorithm is used to eliminate the mismatching pair. The reliability and accuracy of the method are verified by experiments.