随机Hough变换实现角点检测

Corner detection algorithm based on random Hough transformation

为实现角点的有效检测,提高检测速度,提出一种基于随机Hough变换的角点检测方法。利用随机Hough变换求取出直线参数;根据角点在Hough空间中的特征,利用反Hough变换的反演原理对参数空间中的峰值进行反变换,定位图像空间中的直线交点;为避免虚假角点,将那些附近不包含任何边缘的交点删除,得到正确的角点。实验结果表明,该方法相对于Harris算法和SUSAN具有更好的准确性、鲁棒性和稳定性,实时性也有一定提高。

To detect corner effectively,a corner detection algorithm based on random Hough transformation was proposed.Firstly,the linear parameters were obtained using randomized Hough transform.Then,according to the characteristic of corner points in the Hough space,the peaks in the parameter space were inversely transformed using inversion principle of inverse Hough transform,and the intersections of lines in the image space were located.Finally,to avoid false corners and get right corners,the intersections containing any edge deletion was deleted.The experimental results show that the proposed algorithm has better accuracy,robustness and stability compared with Harris algorithm and the SUSAN,and real-time performance can be improved.