A Novel Image Mosaic Method Based on Improved ORB and its Application in Police-UAV
Shuai Wang, Yingying Zhang, Wenshuang Wang, Yanqing Zhao, Shiwei Zhu · 2018
Images acquired by police-UAV (Unmanned Aerial Vehicle) for reconnaissance and forensics need to be stitched in real-time. Image registration affects the quality of image mosaic. For the UAV aerial images with high resolution and rich information, the feature points detected by ORB (oriented FAST and rotated BRIEF) algorithm are unevenly distributed and easy to cluster, which will influence the matching rate and efficiency during image registration. In order to solve the problem, this paper proposes a novel image mosaic method based on improved ORB algorithm. First, a mask is constructed in the image to be registered, and an improved ORB algorithm is adopted to detect and describe feature points. Then, the feature points are matched by Hamming distance and the matched pairs are purified by the Progressive Sample Consensus (PROSAC) algorithm. Finally, the max-flow min-cut algorithm is utilized to determine the best seam-line and the improved Laplacian fusion algorithm is used to eliminate the ghosting, while realizing the seamless splicing of the UAV images. The experimental results show that the proposed method can obtain higher matching rate and efficiency for images with changes of scale, rotation, blur, viewpoint and illumination, and also have good performance in eliminating stitching seam and ghosting.