The paper introduces the HomView-MOT framework to tackle the complexities of multi-object tracking in moving UAV scenarios. It leverages Fast Homography Estimation for efficient computation, Homographic Matching Filter for accurate IoU associations, and View-Centric ID Learning for robust object tracking. The proposed approach outperforms existing methods on VisDrone and UAVDT datasets, showcasing state-of-the-art performance. By integrating innovative techniques, HomView-MOT offers a comprehensive solution for challenging tracking environments.
他の言語に翻訳
原文コンテンツから
arxiv.org
深掘り質問