Feature Matching Method of Oblique Images Based on Geometric Constraints
A feature point and line hierarchical matching method is proposed,suitable for oblique images to solve the challenges of large angle in view transformation,a few matches due to repeated texture,and low matching accuracy.Firstly,the line features of images derive from the line extraction(detection)algorithm(LineSegmentDector),follow constraints to pair,and construct line pair regions to match the improved SIFT feature descriptor.Secondly,after RANSAC algorithm eliminates mismatches,the epipolar constraint acts upon the initial matching results.Then,the obtained lines correct the local image,and the corrected local image uses SIFT matching,which contributes to calculating the original image reversely.The obtained matching points are used to globally correct the oblique image,and the feature points are matched;the grid-based motion statistics(GMS)algorithm eliminates the mismatches;the matching results go through reverse calculation and return to the original image.The line matching results and the point expanding matching results combine into final results,showing that the matching accuracy of the proposed method is close to that of ASIFT,but the number of matching is 1-3 times it.
万方数据
维普
