A Coarse-to-fine Method for High-Resolution Satellite Imagery Matching

Abstract

Image matching is an essential and difficult task in digital photogrammetryand computer vision. This paper presents a triangulation-based hierarchical image matching method for high-resolution satellite imagery. It uses a coarse-to-fine hierarchical strategy with severalimage matching algorithms which combines the matchingresults of the feature points, grid points. Thus, it canprovide dense, precise, and reliable results. First, coefficients of global perspective transformation between images are extracted using the SIFT algorithm with the RANSAC approach, andthe searching image is transformed to the reference coordinate system using these coefficients. Then, an area-based matching is carried out between the reference and the transformed searching images and the matching results are obtained through an inverse transform to return to the original image. For the hierarchical image matching, the parallaxesfrom the previous level can tansfer to the next level with the triangle constraints. At last, in order to eliminate blunders from the matching result, outliers are detected based on the smooth constraint of parallax and 3-D plane fitting of object points. Besides, bidirectional matching method is also adopted. Experiments using Worldview-1 and ALOS images show that the proposed method has the capacity to generating reliable and dense matching results for surface reconstruction from high-resolution satellite imagery.