摘要
角点检测是图像处理和机器视觉领域中基本而重要的问题。提出了一个基于Hough变换的角点检测算法,通过遍历地考察边缘图像中各像素点的一个邻域窗口,利用Hough变换寻找以窗口中心点为出发点的直线段,并通过这些直线段的数量和相互之间的夹角判断当前中心点是否是角点。针对固定的窗口,使用了模板预先计算及环形缓冲区等技术提高Hough变换及图像遍历的效率。在实际图像库上的对比实验表明提出的算法具有高的角点检测率,最低的角点误检率和角点位置误差,同时相比同样可得到构成角点各边的角度的方法.提出的算法同样具有更高的检测率和精度。对于时间性能要求不高的角点检测应用场合,提出的算法比已有算法具有更优的整体检测性能。
Corner detection is a basic and important issue in the area of image processing and machine vision. A corner detection algorithm based on Hough transform was proposed. The neighboring window of each pixel in an edge image is investigated iteratively, and Hough transform is applied to find the straight lines from the center of the window. The number of these lines and the angles formed by each pair of adjacent lines are inspected to determine whether the center pixel of the current window is a corner point. For a window with given size, template computable a priori and circular buffer are utilized to boost the efficiency of the Hough transform and the image pixel traversal process. Comparison experiments on a real image base show that the proposed method gives high corner detection rate. The proposed algorithm also outperforms other methods in low false positive rate of the detection and the corner position errors. Compared with the algorithms giving the arm directions of the detected corners, the novel method also shows a higher arm detection rate and accuracy. For applications that are not time-critical, the proposed algorithm gives better overall detection performance than other algorithms do.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2008年第11期2424-2429,共6页
Chinese Journal of Scientific Instrument
