Abstract:Due to the traditional SIFT algorithm, there are many problems such as slow operation, mis-matching and com.putational complexity in UAV remote sensing image splicing, which can not meet the real-time requirements of remotesensing image processing, and there are exposure differences between the collected images. In this case, the direct super.position and splicing may cause ghosting misalignment at the boundary. In this paper, an improved SURF algorithm and fu.sion algorithm for remote sensing image mosaic of drones were proposed. Firstly, in the feature detection phase, the SURFalgorithm and the Harris corner detection algorithm were combined to obtain the feature points and feature descriptors ofthe image quickly. In the feature matching stage, it was divided into two steps, whinch were rough matching and fine match.ing. The rough matching of the feature points between the stitched images and the fine matching of the mismatched pointswere eliminated by using the RANSAC algorithm. In the image fusion stage, a weighted average algorithm based on dis.tance was used for image fusion. The final experiment showed that the processing speed of the proposed algorithm was im.proved by nearly 5 times compared with the traditional SUFT algorithm. Compared with other improved algorithms, thematching accuracy was also improved, and this algorithm could effectively improve the quality of the image mosaic. The ef.fect solved the problem that the splicing marks, ghosts, dislocations and other phenomena might occur.
