Global and local geometric constrained feature matching for high resolution remote sensing images

Abstract

Feature matching is the foundation of many large-scale vision tasks, such as 3D modeling and simultaneous localization and mapping (SLAM). There are few feature matching algorithms based on geometric constraints. In this paper, we combine global and local geometric constraints. We use the grid-based motion statistics (GMS) feature matching algorithm as the global geometric constraint to obtain high probability correct matching areas and feature matches and then use local geometric constraints to predict and obtain the final matching results. This algorithm can meet the high precision and high efficiency requirements of high resolution remote sensing images. The experiments show that the proposed algorithm can obtain more robust and precise results than the existing feature matching algorithms based on feature descriptions, such as SURF and ORB. Additionally, our algorithm can also be extended to other types of high-resolution images.