基于多尺度局部结构约束的特征点匹配算法  

Multi-scale Local Structure Consistency Based Feature Matching Algorithm

作　　者：刘朝霞[1] 邵峰 李烨 LIU Zhao-xia;SHAO Feng;LI Ye(School of Software,Dalian University of Foreign Languages,Dalian 116044,China;School of Computer and Information Technology,Liaoning Normal University,Dalian 116081,China;Department of Planning and Development,Dalian Bigdata Center,Dalian 116033,China)

机构地区：[1]大连外国语大学软件学院,辽宁大连116044 [2]辽宁师范大学计算机与信息技术学院,辽宁大连116081 [3]大连市大数据中心发展规划部,辽宁大连116033

出　　处：《数学的实践与认识》2023年第10期160-168,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(61806038);辽宁省教育厅科学研究重点项目(2019JYT01);大连外国语大学科研创新团队项目(2017CXTD01)。

摘　　要：对于含有重复模式较多、低重叠区域的图像,特征匹配时会存在离群点较多及伪同构等问题,从而影响匹配的精度.为了提高匹配算法的精度和鲁棒性,本文提出了一种基于多尺度局部结构相似性的特征点匹配算法(MLSS).首先定义局部结构相似性度量方法,即多尺度近邻结构相似性及多尺度局部拓扑结构相似性基于局部结构相似性,构造特征匹配模型,并采用确定性退火思想去除离群点,实现特征点集的精确和快速的匹配.使用Daisy数据集、VGG数据集和航空遥感数据集中的22组典型数据对算法的性能进行验证,匹配结果显示,该算法的精度和鲁棒性优于其他四种经典的特征匹配方法.For the images with large repeated patterns and low overlapping areas,there are many outliers and local pseudo isomorphic structures in feature matching,which decrease the matching accuracy.In order to improve the accuracy and robustness of the matching algorithm,a feature matching algorithm based on multi-scale local structure similarity(MLSS)consistency is proposed.In this algorithm,to evaluate the local structure similarity,the multiscale neighborhood structure similarity and multi-scale local topological structure similarity between the corresponding points are defined.Based on the local structure similarity,the feature matching model is constructed and the mismatches are removed.The performance of the algorithm is verified by 22 images from Daisy dataset,VGG dataset and aerial remote sensing dataset.The matching results show that the accuracy and robustness of the MLSS are better than that of other four typical feature matching methods.

关 键 词：特征匹配 局部结构 多尺度近邻结构 多尺度拓扑结构 匹配模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]