摘要
针对SIFT图像配准算法存在配准精度低的问题,提出了基于SURF的图像配准改进算法。该算法在特征点提取之前,对图像进行双边滤波,减少错误来源,特征匹配初始阶段使用自适应阈值约束代替传统固定阈值,减少最近邻域与次近邻域之比对匹配结果的影响,加入肯德尔系数约束对匹配对提纯,提高配准精度,最后通过RANSAC算法和LSM迭代求解,进行结果处理。实验结果表明,改进的SURF算法在减少配准时间的基础上提高了正确匹配率。
Image registration refers to the alignment of two or more images of the same target in space.In view of the low registration accuracy of SURF image registration algorithm,we proposed an improved algorithm of image registration based on SURF.Firstly,before the extraction of feature points,bilateral filtering was performed to reduce the sources of errors,in the initial stage of feature matching,instead of traditional fixed thresholds,the adaptive threshold constraints became so widespread,which reduced the effect of matching results between nearest neighbor and second nearest neighbor,then the Kendall coefficient constraint pair was added to purify the matching pair to improve the registration accuracy,Finally,through RANSAC algorithm and LSM iteration to get the results processed.Experimental results showed that the improved SURF algorithm promoted the correct matching rate on the basis of reducing registration time.
作者
袁丽英
刘佳
王飞越
YUAN Liying;LIU Jia;WANG Feiyue(Harbin University of Science and Technology,Harbin 150080,China)
出处
《探测与控制学报》
CSCD
北大核心
2020年第2期65-70,78,共7页
Journal of Detection & Control
基金
国家自然科学基金项目资助(61305001)。
关键词
图像配准
SURF算法
双边滤波
肯德尔系数
image registration
SURF algorithm
bilateral filtering
Kendall coefficient