Abstract: A ship detection system based on YOLOv5 and video stabilization is designed and implemented to effectively solve the problems of inflexible deployment, insufficient detection accuracy, and poor real-time performance in traditional sea surface detection solutions. The system consists of a front-end user operation module, a back-end service module, and a sea target detection module, and is deployed on the Jetson AGX hardware platform. The system is designed and implemented in detail, and after integrating the YOLOv5 object detection algorithm and digital video stabilization algorithm, it can accurately identify and remotely monitor sea targets using edge-based unmanned ship-mounted devices, and provide remote control functions. Through experiments, the system has good ship recognition performance, and after optimizing the captured video with stabilization, the recognition accuracy under windy and wavy conditions can be significantly improved. At the same time, according to on-site actual test results, the system provides users with centralized control and management functions through an integrated operation platform, and can achieve a video frame rate of 45 frame/s and a playback delay of less than 2 s, meeting real-time requirements.