Abstract
Image matching is one of the most important tasks in Unmanned Arial Vehicles (UAV) photogrammetry applications. The number and distribution of extracted keypoints play an essential role in the reliability and accuracy of image matching and orientation results. Conventional detectors generally produce too many redundant keypoints. In this paper, we study the effect of applying various information content criteria to keypoint selection tasks. For this reason, the quality measures of entropy, spatial saliency and texture coefficient are used to select keypoints extracted using SIFT, SURF, MSER and BRISK operators. Experiments are conducted using several synthetic and real UAV image pairs. Results show that the keypoint selection methods perform differently based on the applied detector and scene type, but in most cases, the precision of the matching results is improved by an average of 15%. In general, it can be said that applying proper keypoint selection techniques can improve the accuracy and efficiency of UAV image matching and orientation results. In addition to the evaluation, a new hybrid keypoint selection is proposed that combines all of the information content criteria discussed in this paper. This new screening method was also compared with those of SIFT, which showed 22% to 40% improvement for the bundle adjustment of UAV images.
Highlights
Recent developments in Unmanned Arial Vehicles (UAVs) have changed them to a lowcost, flexible and feasible acquisition system for many photogrammetry and remote sensing applications [1,2,3]
The quality of products for UAV photogrammetry depends primarily on the accuracy and completeness of keypoints extracted from the images
The number and distribution of the extracted keypoints play a vital role in the reliability and accuracy of the image matching results, which is a very critical factor in high-resolution UAV photogrammetry bundle adjustment [16]
Summary
Recent developments in Unmanned Arial Vehicles (UAVs) have changed them to a lowcost, flexible and feasible acquisition system for many photogrammetry and remote sensing applications [1,2,3]. To the best of authors knowledge, learning-based keypoint detectors and descriptors start to be available [37,38,39,40], most researchers have focused mainly on developing handcrafted new methods/criteria for keypoint detection and/or description They usually evaluate the overall accuracy of points produced by different detectors. Detectors used for the production of initial keypoints are those which are well known for the extraction of scale and affine invariant features and include scale-invariant feature transform (SIFT) [6], speeded up robust features (SURF) [4], binary robust invariant scalable keypoints (BRISK) [5] and maximally stable extremal regions (MSER) [41] These methods are used and discussed in different previous studies; we can compare our results with those reported in previous studies. A proposal for a hybrid keypoint selection algorithm is presented in Section 7, before concluding the paper with final remarks and suggestion for future studies
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
