基于RFM 的高分辨率卫星遥感影像自动匹配研究[J].pdfVIP

39 6 Vol. 39, No. 6 2010 12 Acta Geodaetica et Ca rtographica Sinica Dec. , 2010 :2010) RFM 季顺平,袁修孝 , 430079 Automatic Matching of High Re olution Satellite Ima ge Ba ed on RFM JI Shunping, YUAN Xiuxiao School of Remote Sen ing a nd Information Engineering, Wuhan Univer ity, Wuha n 430049, China Ab tract : A ma tching method for high re olution atellite image ba ed on RFM i pre ented. Fir tly, the RFM pa rameter a re u ed to predict the init ial pa rallax of corre ponding point and the prediction a ccuracy i analyzed. Secondly, the approximate epipolar equation i con t ructed ba ed on proj ection tracking a nd it a ccuracy i analyzed. Thirdly, approximate 1D ima ge ma tching i executed on pyra mid ima ge and lea t qua re ma tching on ba e ima ge . At la t RANSAC i imbedded to eliminate mi ma tching point a nd matching re ult a re obtained. Te t re ult verified the method more robu t and with higher matching rate, compared to 2Dgray correlat ion met hod and the popula r SIFT ma tching method, a nd the method preferably olved the que tion of high re olution atellite image matching with different tereo model, different time and la rge rota tion image . Key word : image matching; epipolar equation; RANSAC; RFM :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 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 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 :影像匹配; 核线方程; RANSAC; RFM : P237 : A : 国家973 计划(2006CB701302) ;武汉大学自主科研项目( 4082004) 1 引 言 , ,, [4] , [3], SIFT , , [ 1] SIFT