基于优化RANSAC算法的二次元快速稳定配准  

Fast and stable method based on optimized RANSAC algorithm for two-dimensional image

作　　者：孙强[1] 叶玉堂[1] 宋昀岑[1] 张静[1] 姚娇[1] 周恋玲[1] 陈伟[1] 

机构地区：[1]电子科技大学,光电信息学院,四川成都610054

出　　处：《计算机工程与设计》2012年第6期2373-2377,共5页Computer Engineering and Design

摘　　要：针对检测零件缺陷的二次元光电检测仪的速度和稳定性要求,提出了一种加入预匹配的快速稳定的RASNSAC匹配方法。首先对零件图像轮廓区域矩分析以及轮廓多边形逼近以及对其CAD标准档进行解析计算得到图像轮廓和CAD图的重心、主轴方向以及特征点集,然后通过重心和主轴对零件图像和CAD图进行预匹配,最后利用RANSAC对预匹配后的特征点集进行快速匹配。实验结果表明,此改进算法显著地提高了图像的匹配速度和稳定性。A fast and stable pre-matched RANSAC match method is presented to meet the two-dimensional optoelectronic detection instrument＇s speed and stability demands.At first,contour moment analysis and contour polygonalization for the part＇ s image,and the part＇s CAD standard file is parsed to get the image contour and CAD file＇s center,principal axis and the feature point set.To increase the match＇s speed and stability,the part image and CAD standard file are pre-matched based on their centers and principal axises.After pre-matched,RANSAC method is used to match the pre-matched two point sets and optimize the result.The experiment shows that this improved RANSAC algorithm increases the image＇s match speed and stability signally.

关 键 词：多边形逼近 矩分析 主轴 预匹配 随机抽样一致性算法 

分 类 号：TN911.73[电子电信—通信与信息系统]