Image matching techniques for vision-based indoor navigation systems: a 3D map-based approach1

Abstract

In this paper, we give an overview of image matching techniques for various vision-based navigation systems: stereo vision, structure from motion and map-based approach. Focused on map-based approach, which generally uses feature-based matching for mapping and localisation, and based on our early developed system, a performance analysis has been carried out and three major problems have been identified: inadequate geo-referencing and imaging geometry for mapping, vulnerability to drastic viewpoint changes and big percentage of mismatches at positioning. We introduce multi-image matching for mapping. By using affine-scale-invariant feature transform for viewpoint changes, the major improvement takes place on the epoch with large viewpoint changes. In order to deal with mismatches that were unable to be removed by Random Sample Consensus (RANSAC), we propose new method that use cross-correlation information to evaluate the quality of homography model and select the proper one. The conducted experiments have proved that such an approach can reduce the chances of mismatches being included by RANSAC and final positioning accuracy can be improved.