An Aerial Image Stitching Algorithm Based on Long-distance Features

Aerial image stitching is very important in obtaining UAV information. The Speeded Up Robust Features Algorithm (SURF) is an image matching method with high robustness. However, the SURF algorithm subjects to the problems of inexact edge positioning and low matching accuracy. In order to obtain high-precision aerial stitching images with high efficiency, an aerial image stitching algorithm based on long-distance features is proposed in this paper. First, the Canny-SURF algorithm is used for feature detection. After that, the Long-Fast Retina Keypoint (L-FREAK) binary symbol is used to describe and match the feature points. Finally, the Random Sample Consensus Algorithm (RANSAC) is used to compute the projection transformation model, and the weighted average fusion algorithm is used to fuse the pixels. Experimental results show that the proposed algorithm is outperform the SIFT, SURF, ORB, and BRISK algorithms. The proposed algorithm has good stitching accuracy, and can stitch aerial images with wide rotation characteristics.