一种基于梯度Hough变换和SVM的圆检测算法  

Circle Detection Algorithm Based on Gradient Hough Transform and SVM

作　　者：程鹏[1] 朱美琳[1] 耿华[2] 

机构地区：[1]南京大学工程管理学院,江苏南京210093 [2]扬州大学信息工程学院,江苏扬州225127

出　　处：《计算机与现代化》2013年第2期22-26,共5页Computer and Modernization

摘　　要：针对复杂背景下的多圆检测,分析传统的方法,并在标准Hough变换的基础上进行算法改进,把一次三维Hough变换转化成两次二维Hough变换的组合。把圆心和半径分开检测,在积累数组的投票中引入梯度信息作为投票权值,同时采用SVM对检测结果进行假圆的识别和去除,提高了检测精度和算法的效率。Hough transform is a well established method for circle detection.However,circle Hough transform is quite time-consuming.In this paper,some improvements based on standard Hough transform are employed to improve the algorithm.The 3D Hough transform is transformed into the combination of two 2D Hough transform.Gradient information is treated as the value of the voting rights in the accumulated array.SVM is used in the recognition of false circle.The experimental results indicate that the proposed approach can improve the detection accuracy and computational efficiency.

关 键 词：哈夫变换 梯度 支持向量机 圆检测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]