基于随机Hough变换改进的快速圆检测算法  

Fast circle detection algorithm based on improved randomized Hough transform

作　　者：朱正伟[1] 宋文浩[1] 焦竹青[1] 郭晓[1] ZHU Zheng-wei,SONG Wen-hao,JIAO Zhu-qing,GUO Xiao(School of Information Science and Engineering,Changzhou University （Wujin Campus）,Changzhou 213164,China)

机构地区：[1]常州大学(武进校区)信息科学与工程学院,江苏常州213164

出　　处：《计算机工程与设计》2018年第7期1978-1983,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(51307010)

摘　　要：针对随机Hough变换中采样随机性带来的大量无效采样和积累导致算法效率过低的问题,对随机Hough变换的基本原理进行研究,提出一种改进的随机Hough变换快速圆检测算法。利用图片边缘细化处理减少噪声因素的影响,通过随机采样一点和采用最小二乘法对该点所在的N×N窗口内的连通域进行圆拟合,得出候选圆以提高采样效率,通过缩小限定验证取值区域减少计算量,进行候选圆的证据累积判断。实验结果对比分析表明了该算法具有更快的检测速度和较好的准确性。Aiming at the problem of low efficiency caused by random sampling and accumulation in randomized Hough transform（RHT）,the basic principle of the random Hough transform was studied,an improved randomized Hough transform based fast circle detection algorithm was proposed.The influence of noise was reduced using the edge refinement of the image.The candidate circle was rounded out by randomly sampling apoint and using the least squares method in the N×N window where the point was located to improve the sampling efficiency.While by reducing and limiting verification region,the amount of computation load for the accumulation of evidence was reduced to determine a candidate circle.Experimental results show that the algorithm has higher detection speed and accuracy.

关 键 词：随机Hough变换(RHT) 圆检测 边缘细化 随机采样 最小二乘法 证据积累 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]