基于RFM的高分辨率卫星遥感影像自动匹配研究  

Automatic Matching of High Resolution Satellite Images Based on RFM

作　　者：季顺平[1] 袁修孝[1] 

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079

出　　处：《测绘学报》2010年第6期592-598,共7页Acta Geodaetica et Cartographica Sinica

基　　金：国家973计划(2006CB701302);武汉大学自主科研项目(4082004)

摘　　要：提出一种基于有理多项式模型(RFM)的高分辨率卫星遥感影像自动匹配方法。首先利用RFM进行高分辨率卫星影像直接定位和同名点预测,并分析预测精度;然后基于投影轨迹建立近似核线方程,并分析核线精度;接着采用金字塔影像策略进行核线约束的近似一维影像匹配,并经最小二乘影像匹配精化匹配结果;最后采用RANSAC算法剔除误匹配点以获取最终的匹配结果。试验结果表明,该方法的匹配成功率和稳定性高于传统的二维灰度匹配方法和现流行的SIFT匹配方法,能够很好地解决高分辨率卫星遥感影像自动匹配中不同成像模式、多时相、大姿态角等情况导致的匹配难题。A matching method for high resolution satellite images based on RFM is presented.Firstly,the RFM parameters are used to predict the initial parallax of corresponding points and the prediction accuracy is analyzed.Secondly,the approximate epipolar equation is constructed based on projection tracking and its accuracy is analyzed.Thirdly,approximate 1D image matching is executed on pyramid images and least square matching on base images.At last RANSAC is imbedded to eliminate mis-matching points and matching results are obtained.Test results verified the method more robust and with higher matching rate,compared to 2D gray correlation method and the popular SIFT matching method,and the method preferably solved the question of high resolution satellite image matching with different stereo model,different time and large rotation images.

关 键 词：影像匹配 核线方程 RANSAC RFM 

分 类 号：P237[天文地球—摄影测量与遥感]