基于SUSAN算子和角点判别因子的目标边缘检测  

Target edge detection based on SUSAN operator and corner discriminant factor

作　　者：吴一全[1,2,3,4] 王凯[1] 

机构地区：[1]南京航空航天大学电子信息工程学院,南京210016 [2]西华大学制造与自动化省高校重点实验室,成都610039 [3]深圳大学深圳市城市轨道交通重点实验室,广东深圳518060 [4]南京财经大学江苏省粮油品质控制及深加工技术重点实验室,南京210046

出　　处：《中国科学院大学学报（中英文）》2016年第1期128-134,共7页Journal of University of Chinese Academy of Sciences

基　　金：国家自然科学基金(60872065);制造与自动化省高校重点实验室开放课题(2014);高速铁路线路工程教育部重点实验室开放基金(2014-HRE-01);江苏省粮油品质控制及深加工技术重点实验室开放基金(LYPK201304);江苏高校优势学科建设工程项目(2012)资助

摘　　要：针对目标区域角点分布密集和背景区域相对稀疏的图像,为了更准确、完整地提取目标区域的边缘,消除背景,提出一种基于SUSAN算子和角点判别因子的目标边缘检测方法.实验结果表明,与Canny方法、改进的非下采样Contourlet模极大值方法和改进的蜂群方法等边缘检测方法相比,本文提出的方法能有效避免背景区域的干扰,精确定位目标区域,所得边缘轮廓连通完整、细节丰富.该方法具有较优的主观视觉效果和较强的抗噪能力,且运行时间较少.This work aims at the images with intensive corners in the target area and sparse corners in the background area. To extract the edges of target area more accurately and more completely and to eliminate the background,a target edge detection method based on SUSAN operator and corner discriminant factor is proposed. First,the corners of image are extracted by SUSAN operator and the isolated noise points in the image are filtered. Then,target corner discriminant factor is defined for elimination of corners in the background area and preservation of corners in the target area. Finally,according to the similar standard of effectiveness,other edge points are detected based on the targetcorner standard and the target edges are obtained. A large number of experimental results show that,compared with Canny method,the improved bee colony method,and the improved non-subsampled contourlet modulus maxima method,the proposed method avoids the interference of background area effectively and locate the target area accurately. The obtained edge profile can be connected and is complete with abundant details. It has better subjective visual effect and stronger anti-noise ability with less running time.

关 键 词：目标边缘检测 角点 SUSAN算子 目标角点判别因子 效益度相近标准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]