图结构在航空遥感图像特征点匹配中的应用  

Application of graph structure in aerial image feature matching

作　　者：刘朝霞[1] 邵峰 景雨[1] 祁瑞华[1] 

机构地区：[1]大连外国语大学软件学院,辽宁大连116044 [2]大连海事大学信息科学技术学院,辽宁大连116026

出　　处：《计算机工程与应用》2018年第1期19-24,共6页Computer Engineering and Applications

基　　金：国家自然科学基金青年基金(No.61201454;No.61501082);国家社科基金一般项目(No.15BYY028);辽宁省高等学校优秀人才支持计划(No.LJQ2014127);辽宁省自然科学基金(No.2015020017;No.20170540230;No.20170540232);2014年度大连外国语大学科研基金(No.2014XJYB12)

摘　　要：为了解决海上航空遥感图像拍摄目标及相机视角变化快,相似特征多导致的匹配不一致问题,并提高匹配算法的效率,提出了一种基于图结构的航空遥感图像特征点匹配算法。该特征匹配过程分为初始匹配和精确匹配两步,首先采用快速特征提取算法SURF进行特征提取并进行初始匹配,然后在精确匹配过程中充分考虑图像的局部空间结构及全局信息,建立K近邻图结构,用K近邻结构差异与点集的变换误差作为匹配的收敛条件,以解决图像目标发生变化而导致的K近邻结构不一致及K近邻结构相同时仍然存在干扰点两个问题,最终实现快速精确的特征匹配。The aerial remote sensing image captured on the sea has the characteristics of dynamicity, monotonous and similar patterns, which may result in mismatches in feature matching. In order to remove the mismatches and improve the efficiency, a feature point matching algorithm based on graph structure is proposed. There are two stages in feature matching process, coarse matching and fine matching. In the coarse matching, Speeded Up Robust Features(SURF)is used to get an initial matching results. In the fine matching, Restricted GTM(RGTM)is proposed. Both local spatial structure and global information are considered. KNN graphs are constructed by the initial points. Then feature matching is converted to graph matching and global transformation error is used to restrict the matching convergence. The experimental results show that the proposed algorithm is more efficient and accurate than GTM.

关 键 词：特征匹配 SURF特征 结构特征 全局变换误差 图结构 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]