Real-time Online UAV Images Mosaic with Robustness to Cumulative Errors
Xing Xia, Bei Wang, Jing Li, Song Li, Yi Gong, Bao Song Deng · 2020
In this paper, a novel real-time online UAV image mosaic algorithm is proposed to meet the real-time, robustness and accuracy requirements of practical application. First, the algorithm adopts GPS and track planning information to locate the adjacent images. Instead of the 8-DOF (degrees of freedom) homography transformation, the 4-DOF similarity transformation is estimated to more robustly register the UAV images and reduce their perspective errors. Second, a new method of local registration is developed to reduce the cumulative errors and calculation costs. Meanwhile, the real-time mosaic process is used to register images and update the panoramic image. Finally, the effectiveness of the proposed approach is evaluated by a group of experiments on 4K images taken by quadrotor drones. According to the comparison of experiments, the proposed algorithm is robust to cumulative errors and obtains high-quality panorama. Additionally, through the analysis on stitching time and practical application, frame-by-frame mosaic synthesis is online updated in real-time.