Design of Autonomous Object Detection System for USV

This paper introduces an autonomous system designed for sea surface objects detection, addressing the needs of Unmanned Surface Vehicles (USVs) involved in tasks like environmental monitoring, sea patrol and so on. The system also uses Generative Adversarial Networks (GANs) for data enhancement, an object detection algorithm based on YOLOv5 with dense connection modules for Infrared target detection and image enhancement for model training. Related tests have been done for validating the algorithm's performance, showing its ability to detect objects across varying weather conditions and lighting scenarios. This system meets the requirements of autonomous obstacle avoidance and object detection for USVs during the mission time. While the preliminary tests indicated a capability index slightly below the dataset results, ongoing investigations seek to elucidate the reasons behind this discrepancy. Nevertheless, the system meets the situational awareness needs of USVs in the project of the author.