文献类型：期刊文献

中文题名：改进SURF和Delaunay三角网在图像匹配中应用

英文题名：Application of improved SURF and Delaunay triangulation in image matching

作者：毛克乐[1]

第一作者：毛克乐

机构：[1]新乡学院现代教育技术中心,河南新乡453003

第一机构：新乡学院

年份：2021

卷号：43

期号：4

起止页码：432-438

中文期刊名：沈阳工业大学学报

外文期刊名：Journal of Shenyang University of Technology

收录：CSTPCD;;Scopus;北大核心:【北大核心2020】；

基金：河南省科技攻关项目(152102210202).

语种：中文

中文关键词：图像匹配;SURF算法;Delaunay三角网;邻近特征点;颜色不变量模型;三角形相似函数;射影不变量;空间变换

外文关键词：image matching;SURF algorithm;Delaunay triangulation;adjacent feature point;color invariant model;triangle similarity function;photographic invariant;spatial transformation

摘要：针对SURF算法中存在较多错误匹配问题,提出一种基于改进SURF和Delaunay三角剖分图像匹配算法.以颜色不变量模型作为SURF的输入,利用邻近特征点之间的关系,解决SURF引起的颜色成分信息丢失和特征点过于密集问题.利用三角形相似函数计算两幅图像中Delaunay三角形相似度大于0.75的三角形,并采用射影不变量执行空间变换处理进行粗匹配和精匹配.结果表明,与当前图像匹配算法相比,该算法具有更好的精度与鲁棒性,提取特征点多且分布均匀.
Aiming at the problems of more mismatches in SURF algorithm,an image matching algorithm based on improved SURF and Delaunay triangulation was proposed.A color invariant model was used as the input of SURF,and the relationship between adjacent feature points was used to solve the problems of color component information loss and extremely dense feature points caused by SURF.The triangle and photographic invariants in two images with the Delaunay triangle similarity greater than 0.75 were calculated by a triangle similarity function to perform spatial transformation for coarse and fine matching.The results show that the as-proposed algorithm has better accuracy and robustness,more feature points and even distribution,compared with the currently available image matching algorithms.