Deep-learning-empowered visual ship detection and tracking: Literature review and future direction

Boxing Zhang, Jingxian Liu, Ryan Wen Liu, Yanhong Huang

Abstract: Visual object detection and tracking is a fundamental task in artificial intelligence and plays a crucial role in various maritime applications, leading to enhanced maritime traffic safety and security. Although significant progress has been made in object detection and tracking in recent years, it still remains a challenge due to the image quality decline and object appearance changes caused by complex maritime traffic environments. This article aims to provide a comprehensive review of the current state of research in the field. We have examined the latest advancements in various aspects and proposed some directions for future research. Firstly, we delve into four main challenges in ship detection and tracking, including environmental variations, ship scale, ship occlusion, and lightweight models. Subsequently, we review the methods employed to address these challenges, and key contents about ship detection and tracking, include datasets, evaluation metrics, experimental results from mainstream methods. Finally, we offer a discussion and some future research directions that warrant attention, which will become potential research effort in the future.