Research on a YOLOv5-based crack detection system for naval ships

With the rapid development of the shipbuilding industry, the efficient detection of cracks on ships has become a key issue to ensure the safety of navigation. This paper proposes a real-time ship crack detection system based on a magnetic wall-climbing robot equipped with a K210 camera and YOLOv5 algorithm, to solve the problems of low efficiency and limited coverage of traditional manual detection. The system realizes flexible movement and stable image acquisition on the surface of complex ship hulls through the magnetic wall climbing robot, and uses the K210 camera for real-time image capture and compression transmission; YOLOv5 algorithm is used to construct a ship crack detection model, and through the training and optimization of the data set of 3,500 diverse scenes, the model accuracy reaches 0.93, and achieves the real-time reasoning capability of 10 FPS. The experimental results show that the system can effectively identify surface cracks on ships, support visualization result output and real-time data recording, with high engineering practicability and detection efficiency, and provide an innovative solution for ship maintenance intelligence.