摘要
为了降低传统椭圆检测方法对于边缘梯度精度和椭圆完整性的过高要求,有效地提取出图像中的多个未知椭圆,该文提出了一种基于对称点搜索的仅需要一维参数空间累加阵列的快速椭圆检测方法。该方法充分利用椭圆的对称性,借助随机获取的对称点组确定候选椭圆的3个待定参数,并引入基于椭圆生成和膨胀匹配的候选椭圆确认方法用于克服椭圆形变对候选椭圆确认造成的影响。实验结果表明该文算法与基于RHT的其他具有代表性的椭圆检测方法相比具有检测速度快,精度高,抵抗椭圆的部分缺失和形变能力强等优点。
To lower the high requirements of both the accuracy of the edge gradient and the integrity of the ellipse, a fast ellipse detection method using only 1-D accumulation array in parameter space based on symmetry points searching is presented to effectively extract multiple unknown ellipses in an image. Three undetermined parameters of the candidate ellipse are determined by the group of the symmetry points which are acquired by random sampling and the validation algorithm based on the ellipse generation and dilation matching is introduced to eliminate the effect of the ellipses' distortion on the validation of the candidate ellipses. The experimental results show that compared with the other typical ellipse detection method based on RHT, the method here has the advantage of higher detection speed and accuracy and strong resistance to ellipse's partial deletion and distortion.
出处
《电子与信息学报》
EI
CSCD
北大核心
2009年第8期1830-1834,共5页
Journal of Electronics & Information Technology
基金
浙江省科技计划优先主题项目(2008C13056)资助课题
关键词
椭圆检测
随机HOUGH变换
膨胀匹配
椭圆生成
Ellipse detection
Randomized Hough Transform (RHT)
Dilation matching
Ellipse generation
作者简介
于海滨:男,1979年生,讲师,研究方向为图像处理与计算机视觉、模式识别、计算机并行处理.
刘敬彪:男,1964年生,教授,主要研究方向为信号处理、多媒体通信、智能仪器仪表、工业自动化等.
刘圆圆:女,1978年生,讲师,研究方向信号与信息处理、图像处理与机器视觉.