基于SIFT宽基线立体影像密集匹配.pdfVIP

40 5 Vol.40, No.5 2011 10 Acta Geodaetica et Cartographica Sinica Oct.,2011 :1001-1595(2011) 05-0537-07 SIFT 1, 2 2 1, 2 杨化超 , 姚国标 , 王永波 1. , 221116; 2. , 221116 Dense Matching for Wide ase-line Stereo Images ased on SIFT 1, 2 2 1, 2 YANG Huachao , YAO Guobiao , WANG Yongbo 1. Key Laboratory for Land Environment Disaster Monitoring of S SM, Xuzhou 221116, China; 2. School of Environmental Spatial Informatics, China University of Mining Technology,Xuzhou 221116, China Abstract :A novel mult-i stage quas-i dense matching algorithmfor wide base-line stereo imagesis introduced based on SIFT a nd dual constra ints of epipola r geometry and homographic mapping. The proposed a lgorithm includes following three stages: The optima l SIFT features withgood spatial distributiona nd large informationcontent are first selected, and matched by using the least squares matching method, then the fundamenta l and homographic matrix can be estimated by using these initia l sparse correspondences with higher precision; For the other SIFT features, the affine tra nsformation para meters between matching windows a re iteratively optimized by using the slope a ngle of correspondent epipola r lines and scale information of SIFT features, a nd affine invaria nt feature descriptors are extracted fromthe corrected matchingwindows, then correspondences can be determined by Euclid dista nce and dua l constraint information; Considering the lower repeatability rate of feature detection for wide base-line stereo images, for the unmatched points extracted from left and right images of stereo pa irs, matching ca n be carried out by adopting two-way searchstrategy from left to right image or from right to left image based on the rapid SSD similarity cost function and affine r