计算机科学
水下
遥感
地质学
噪音(视频)
跟踪(教育)
物理
联轴节(管道)
人工智能
跟踪系统
作者
Xiaoyan Liu,Junfeng Fan,Yaming Ou,Chao Zhou,Shaoyang Sun,Zeng‐Guang Hou
出处
期刊:IEEE-ASME Transactions on Mechatronics
[Institute of Electrical and Electronics Engineers]
日期:2026-07-29
卷期号:31 (4): 4084-4095
标识
DOI:10.1109/tmech.2026.3715067
摘要
Accuratelocalization of underwater robots is a prerequisite for autonomous operation. However, existing visual systems often fail in complex underwater environments, including low light and dynamic light. Event cameras offer a promising solution due to their superior adaptability to extreme lighting. Therefore, we propose an underwater self-localization system AEVINS to achieve robust localization in complex underwater environments. First, an adaptive slicing and motion compensation method is proposed, which integrates an adaptive slicing strategy based on event stream characteristics and a motion compensation algorithm independent of the inertial measurement unit (IMU) to generate clearer event frames. Subsequently, an acoustic-event-visual-inertial tightly coupled localization method is proposed. A joint optimization problem is constructed by using reprojection error terms of event and standard frames, along with inertial and Doppler velocity log error terms to enhance system robustness. Finally, a multisensor robotic platform was developed, and comprehensive experiments were conducted in various complex environments, including low light and dynamic light pool scenes, as well as turbid, uneven-textured, and weakly textured sea scenes. Experimental results indicate that the proposed method achieves superior accuracy and robustness compared to state-of-the-art underwater visual fusion and event-based localization methods.
科研通智能强力驱动
Strongly Powered by AbleSci AI