Using Information Content to Select Keypoints for UAV Image Matching

Image matching is one of the most important tasks in Unmanned Arial Vehicles (UAV) photogrammetry applications. The number and distribution extracted keypoints play an essential role reliability accuracy image orientation results. Conventional detectors generally produce too many redundant keypoints. In this paper, we study effect applying various information content criteria to keypoint selection tasks. For reason, quality measures entropy, spatial saliency texture coefficient are used select using SIFT, SURF, MSER BRISK operators. Experiments conducted several synthetic real UAV pairs. Results show that methods perform differently based on applied detector scene type, but cases, precision results improved by average 15%. general, it can be said proper techniques improve efficiency addition evaluation, a new hybrid proposed combines all discussed paper. This screening method was also compared with those which showed 22% 40% improvement for bundle adjustment images.