Analysis of the Current State, Problems and Development Trends of Computer Vision in Unmanned Surface Vehicle Systems

Abstract

This article reviews modern approaches to the application of computer vision in autonomous unmanned surface vehicles (USVs). Key challenges associated with visual perception in aquatic environments are analyzed, including illumination instability, surface reflections, waveinduced disturbances, and the limited computational resources of onboard systems. State-of-the-art algorithms for object detection, classification, and segmentation on water surfaces are examined, along with Edge-Cloud architectural approaches for visual data processing. Based on a review of the literature, promising directions for the advancement of computer vision are identified with the aim of enhancing the autonomy, navigational safety, and operational efficiency of surface robotic platforms. Special attention is given to the Edge AI paradigm and the potential of compact onboard computing platforms (edge devices) to overcome the resource constraints inherent to unmanned surface vehicles.