一种适用于图像测量的角点检测算法

摘要

A sub-pixel corner detection algorithm based on edge fitting is proposed in this paper,according to the geometric characteristics that a corner must be the intersection of two edges or more.First,image regions that cover the edges forming corners are obtained,and their equations are obtained based on sub-pixel edge detection algorithm in these regions.Second,the coordinates of four intersections formed by two adjacent lines are solved according to the equations.Finally,the subpixel coordinate of the corner are solved by solving the mean of four.By comparing the algorithm proposed by Bouguet,the mean absolute errors of measurement values with this corner detection algorithm are less,when the visual effect and projection error are close,which probe that this algorithm is more suitable for high-accuracy vision measurement.%根据角点一定是边缘点且是2条边缘或多条边缘交点的几何特征,提出了一种基于亚像素边缘拟合的角点提取算法.首先,获得形成角点的边缘所处的图像区域,在该区域中利用亚像素边缘检测算法求得边缘的方程,然后利用相邻2条直线方程分别求交点,最终取均值求出角点的亚像素坐标.通过与Bouguet算法比较,发现在视觉比较、投影误差相近的情况下,本文角点检测算法获得的尺寸测量值的平均绝对误差要小于Bouguet算法,证明本算法更适用于高精度的图像测量。