基于改进轮廓提取的Hough变换椭圆检测方法  

A Hough transform ellipse detection method based on improved contour extraction

作　　者：翟永立 丁雷[1,2] 裴浩东[1,2] ZHAI Yongli;DING Lei;PEI Haodong(Key Laboratory of Infrared Detection and Imaging Technology,Chinese Academy of Sciences,Shanghai 200083,China;Shanghai Institute of Technical Physics,Chinese Academy of Sciences,Shanghai 200083,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院红外探测与成像技术重点实验室,上海200083 [2]中国科学院上海技术物理研究所,上海200083 [3]中国科学院大学,北京100049

出　　处：《现代电子技术》2019年第6期34-37,共4页Modern Electronics Technique

基　　金：国家"863"项目(2015AA7046604)~~

摘　　要：分析椭圆目标在成像平面上的特点,提出一种基于改进轮廓提取的Hough变换椭圆检测方法,可以有效地在图像中提取出椭圆的轮廓。对边缘检测后的图像进行边缘点分类,去除交叉点和孤立点,保存端点;根据端点,提取图像中的非封闭轮廓,再扫描图像,提取剩余的封闭轮廓;针对随机Hough变换,根据每条轮廓的长度判断该轮廓需要采样计算的次数,在每个循环的内部,设定阈值来判断每组椭圆参数的统计值,用以跳出循环。实验结果表明,与普通随机Hough变换相比,该算法的内存占用量与计算量大大减少。The characteristic of the elliptical object on the imaging plane is analyzed,and a Hough transform ellipse detec-tion method based on improved contour extraction is proposed,which can effectively extract the outline of the ellipse in the im-age. The edge points of images are classified after edge detection,so as to remove cross points and isolated points,and save endpoints. The non-closed contour in the image is extracted according to the endpoints. The remaining closed contour is extract-ed by scanning images. In allusion to the random Hough transform,the times of sampling and calculation are determined for the contour according to the length of each contour. A threshold value is set to judge the statistical value of each group of elliptic pa-rameters inside each loop,so as to jump out of the loop. The experimental results show that the memory consumption and calcu-lation amount of the algorithm are greatly reduced in comparison with the ordinary random Hough transform.

关 键 词：椭圆目标 边缘点分类 检测方法 轮廓提取 随机HOUGH变换 采样次数 

分 类 号：TN911.23-34[电子电信—通信与信息系统] TP391.9[电子电信—信息与通信工程]