UAV aerial image mosaic algorithm based on FAST-Tomasi feature and Delaunay triangulation
Hui Zhao, Yaxin Du, Hongjun Wang, Youjun Yue · 2020
In view of the current inability to meet the requirements of fast speed and strong stability in feature extraction, and the slow speed in feature point elimination, this paper proposes an UAV remote sensing image mosaic algorithm based on FAST-Tomasi features and Delaunay triangulation. Firstly, FAST feature detection algorithm is used to accelerate Shi-Tomasi algorithm to obtain fast and accurate feature points, the feature points are described by BRIEF operator, and the phase is conducted by hamming distance. Then, Delaunay triangulation network is constructed according to the initial matching points, and the similarity between the triangles in the two images to be stitched is calculated. The triangle pair with large similarity is the matching triangle pair, so the corresponding accurate matching points are obtained. Then, the improved SPHP algorithm is used for registration. Finally, the weighted algorithm is used to fuse the image. Experimental results indicate that the algorithm retains more correct matching points, the accuracy of eliminating false matching points is higher, and the matching speed is significantly improved.