Deep learning enhanced object detection for unmanned surface vehicles across diverse sensor types: a state-of-the-art review
Tian Xu, Zihan Yang, Qian Ye, Yong Bai, Mingfeng Huang, Fang Wang, Jie Bai, Liang Zhao · Ships and Offshore Structures · 2026
Deep learning plays a crucial role for object detection for unmanned surface vehicles (USVs), enabling safe navigation under complex ocean environments. Due to the intricate marine condition and diverse sizes of ship targets, establishing robust and reliable perception techniques remains a key challenge. This paper presents a comprehensive overview of recent deep learning approaches for USV object detection. We systematically analyze the current research landscape, and summarize detection porgress across various onboard sensors, including RGB/thermal cameras, LiDAR, radar, and data fusion methods. We review the commonly used marine datasets, covering dataset size, target detection types, labels, application scenarios. Furthermore, we discuss the emerging applications and provide potential future research initiatives. The analysis highlights that improving robustness, multi-scale detection, efficiency, sensor integration, multi-task ability, and dataset diversity holds great promise. This study aims to provide the theoretical fundamentals and recent research insights for beginners and professionals in autonomous surface vehicles.