Real-Time UAV Multi-Video Image Stitching Framework for Emergency

In applications such as emergency rescue, security surveillance, and environmental monitoring, unmanned aerial vehicles (UAVs) have emerged as essential tools for acquiring real-time video information due to their flexibility and high mobility. However, UAV-captured footage is often discrete, partial, and insufficient for directly providing a complete global view. Traditional image stitching methods exhibit significant limitations in terms of real-time performance, stitching accuracy, and error control. To address these challenges, this paper presents a real-time multi-video stitching framework for UAVs, termed ReaMIS, based on existing research. The proposed ReaMIS framework aims to enhance the efficiency and accuracy of multi-view UAV video stitching by integrating visual simultaneous localization and mapping (VSLAM) with global navigation satellite system (GNSS) data for camera pose estimation. This integration effectively mitigates drift errors caused by prolonged flight operations. Additionally, ReaMIS leverages sparse point cloud estimation to reconstruct surface information, thereby improving geometric consistency in the stitched images. Furthermore, a perspective correction and global stitching strategy is introduced, optimizing image fusion by minimizing elevation variance to enhance stitching stability. In the experimental evaluation, we validate the proposed framework using publicly available datasets (UAV123, VisDrone) as well as self-collected UAV videos. Experimental results demonstrate that, compared with conventional stitching approaches such as SIFT, ORB, and APAP, ReaMIS achieves superior stitching accuracy, computational efficiency, and error control. Specifically, in high-resolution video stitching scenarios, the proposed framework improves frame rates by approximately 40%. This study provides technical support for real-time information processing in UAV-assisted emergency response and offers insights for future research on multi-UAV collaborative stitching and lightweight on-device stitching algorithm optimization.