摘要
在双目视觉特征提取和匹配方法研究中,为提高匹配率、角点定位精确性以及算法的时间复杂度三项指标,文中提出了一种新的角点提取和匹配融合算法.对提取的Harris角点进行零均值归一化互相关系数匹配,对SIFT角点进行最近邻匹配.针对可预见的误匹配,采用RANSAC算法剔除误匹配点.仿真实验表明:RANSAC算法可以剔除绝大部分的误匹配点,SIFT角点融合最近邻匹配较Harris角点融合零均值归一化互相关系数匹配效果理想.
In research of binocular vision feature extraction and match algorithm,this paper proposes a kind of new corner detection and matching fusion algorithm on purposes of improving such three indicators as matching rates,corner locating accuracy and time complexity.For extracted Harris corners,the algorithm can cope with correlation coefficient matching after zero-means normalization.And SIFT corners can be treated with nearest neighbor matching.The RANSAC algorithm can be applied to rid mismatching points in case of foreseeable mismatching.After simulation,RANSAC can eliminate most mismatching points and SIFT-NN performs better than Harris zero-means normalization correlation coefficient matching.
出处
《微电子学与计算机》
CSCD
北大核心
2013年第7期120-123,共4页
Microelectronics & Computer
基金
国家自然科学基金(AA200904215)
