High-precision Image Mosaic Algorithm Based on Adaptive Homography Transform

The Unmanned Aerial Vehicle (UAV) image mosaic is a research hotspot of digital image processing in recent years, which has extensively application prospect. The existing algorithms are sensitive to aerial images mosaic with large scale and perspective differences, and have the disadvantages of poor accuracy and image distortion, which can not obtain the global optimal mosaic effect. An aerial images mosaic method based on the adaptive homography matrix is proposed to realize high-precision mosaic of multi-scale and multi-view aerial images. The feature points of different UAVs aerial images are constructed by the Speeded Up Robust Features (SURF) algorithm. Fast Library for Approximate Nearest Neighbors (FLANN) is used to realize rough matching of feature points. The Random Sample Consensus (RANSAC) algorithm is used to eliminate the mismatched points to achieve the fine matching, and the position-dependent adaptive homography transform model is constructed to complete matching. Finally, the weighted average method is implemented to complete the fusion for the purpose of seamless image mosaic. The performance of the mosaicing algorithm based on adaptive homography transform is verified on a practical aerial dataset, and the results are compared with the related algorithms. Experimental results show that the proposed algorithm satisfies the requirements of high-precision mosaic, and has high efficiency.