一种改进的快速SUSAN棋盘格角点检测算法  

An Improved Fast SUSAN Chessboard Corner Detection Algorithm

作　　者：夏海英[1] 刘伟涛 朱勇建 

机构地区：[1]广西师范大学电子工程学院,广西桂林541004

出　　处：《广西师范大学学报（自然科学版）》2018年第1期44-52,共9页Journal of Guangxi Normal University:Natural Science Edition

基　　金：国家自然科学基金(51775230;61275110);广西研究生教育创新计划项目(XYCSZ2017052);桂林市科学研究与技术开发项目(2016010604;20170104-2);广西师范大学重点项目(2015ZD004);广西高等教育本科教学改革工程项目(2017JGB148;2017XJGQ25)

摘　　要：针对传统SUSAN角点检测算法不能区分棋盘格内角点和边缘点,以及因曝光过度导致的角点分离问题,本文提出改进的SUSAN角点检测算法。改进算法求取2组不相邻特征方向3像素灰度和差值,配合门限阈值和差阈值,提取角点集合;利用角点集合的对称分布特性确定角点集合的质心坐标作为此位置角点的坐标,坐标值可直接达到亚像素精度。实验表明:本文提出的改进的SUSAN角点检测算法可以有效、快速地提取角点,重投影误差在0.3个像素以内,可用于相机标定。Since the traditional SUSAN corner detection algorithm can not distinguish inner corners and to solve the edge corners of chessboard, and to solve the problem of corner separation caused by excessive exposure, the improved SUSAN corner detection algorithm is proposed. The improved algorithm can obtain the difference of 3 pixels gray sum between two groups of nonadjacent characteristic directions, combining the threshold and difference threshold to extract the corner set. Moreover, using the symmetric distribution of the corner set to determine the eentroid coordinates of the corner set as the coordinates of the corner of the position, and the coordinate values can he directly sub- pixel accuracy. Experiments show that the proposed algorithm can extract corners effectively and quickly,and the reprojeetion error is within 0.3 pixels. The extraction results can be used for camera calibration.

关 键 词：角点检测 棋盘格 相机标定 亚像素 SUSAN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]