基于链码的随机Hough变换圆形检测方法

该文提出基于链码的随机Hough变曲变换圆形检测方法.该方法首先采用改进的自适应性Canny边缘检测方法提取边缘,通过LoG滤波抑制检测中噪声和光照变化的干扰,并采用大津法实现Canny算法中双门限的自动选取,提高Canny算法的自适应性.然后通过八邻域跟踪算法构建边缘链码,采用邻域均值投票,可有效减少零散或小尺度边缘的冗余投票,避免虚假圆生成.同时根据链码长度设定优先级,减少重复检测.实际在线检测结果表明该方法具有良好的准确性,低漏检和误检率,在噪声、光照变化下具有良好的鲁棒性.

A Chain-code Based Prior Randomized Hough Transform Method for Circle Detection

This paper proposes a chain-code based prior randomized Hough transform method for circle detection.At first,an improved adaptive Canny edge detection algorithm is adopted to extract edges,suppressing the interference of noise and lighting change by LoG filtering,achieving automatic selection of two threshold by Otsu that improves self-adaptive of Canny algorithm.Then,eight-neighborhood chain code tracking algorithm making up strings of edge chain code and neighbor-mean vote can effectively reduce votes of scatted or small-scale edges,avoid pseudo circles.At the same time,setting priorities according to the length of chain code can reduce reduplicate detection.Actual online detection results show that this method has good accuracy,low miss and false detection rate.And it has well robustness under the interference of noise and lighting change.