计算机工程

• 图形图像处理 • 上一篇    下一篇

基于采样约束的随机Hough变换圆检测算法

陈树越,李建森,黄萍,徐来   

  1. (常州大学 信息科学与工程学院,江苏 常州 213164)
  • 收稿日期:2016-07-22 出版日期:2017-10-15 发布日期:2017-10-15
  • 作者简介:陈树越(1963—),男,教授、博士,主研方向为图像检测;李建森、黄萍、徐来,硕士研究生。
  • 基金项目:
    国家自然科学基金“温度对X射线与物质相互作用的影响及其测量研究”(51176016);江苏省产学研联合创新基金“高频条件下电子变压器参数测试的关键技术研究与开发”(BY2014037-11)。

Randomized Hough Transform Circle Detection Algorithm Based on Sampling Constraint

CHEN Shuyue,LI Jiansen,HUANG Ping,XU Lai   

  1. (School of Information Science and Engineering,Changzhou University,Changzhou,Jiangsu 213164,China)
  • Received:2016-07-22 Online:2017-10-15 Published:2017-10-15

摘要: 为减少图像中的圆检测时间并提高圆检测精度,提出一种随机Hough变换圆检测算法。以随机采样的一点作为圆直径的一个端点,用约束候选圆直径范围的方法确定候选圆,再将候选圆n等分,对每个方向进行证据积累以判断是否为真圆,通过聚类算法剔除错误样本点,对真圆进行最小二乘拟合以得到更准确的圆参数,解决由于随机采样三点而造成的大量无效采样累积问题。实验结果表明,与RHT算法、ARHT算法和BRHT算法相比,该算法的检测速度更快、检测精度更高。

关键词: 采样约束, 随机Hough变换, 聚类算法, 最小二乘法, 圆检测, 无效累积

Abstract: In order to reduce the detection time and improve the accuracy of circle detection in image,a randomized Hough transform circle detection algorithm is proposed.A random sampling point is selected as an endpoint of the circle diameter and the candidate circle is determined by constraining the range of the candidate circle diameter.Then the candidate circle is divided equally into n parts and the true circle can be judged by the evidence accumulation at each direction.The error sampling points are eliminated through clustering algorithm and more accurate circle parameters can be obtained by least squares fitting on the true circle.This can solve the problem of large amounts of invalid sampling accumulation due to random three-point sampling.Experimental results show that compared with RHT algorithm,ARHT algorithm and BRHT algorithm,the proposed algorithm has faster detection speed and higher detection accuracy.

Key words: sampling constraint, randomized Hough transform, clustering algorithm, least squares method, circle detection, invalid accumulation

中图分类号: