Abstract: Smelting is an important component of national economic development. In future development, smelters will play a crucial role. The improvement and development of smelting technology enable us to more effectively utilize the resources on Earth, while also promoting technological progress in human society. In the production and operation process of smelting enterprises, the crude copper sampling operation is an important production link directly related to the product quality of the smelting and blowing process. The research on automatic sampling systems is particularly important for the production and management level of fine smelting technology. This paper is based on the intelligent crude copper automatic sampling and detection system of YOLOv5, which is developed on the basis of machine vision technology and deep learning framework based on YOLOv5. The configurable SPP layer method and optimized NMS non maximum suppression algorithm are proposed to provide visual recognition accuracy for crude copper. The NMS optimization algorithm is optimized through CPU operation, and the algorithm is simplified and optimized to improve its computational efficiency, solving the problem of accessing and processing multiple video data on a single device has significant implications for practical applications on site. This paper aims to solve the problem of automatic sampling of crude copper in smelters, achieve unmanned and intelligent operation of the crude copper sampling process, improve the efficiency of crude copper sampling, reduce personnel labor intensity, and achieve the goals of reducing personnel and increasing efficiency. This system helps to stabilize the production operation rate, improve the efficiency of crude copper sampling, avoid the risk of cheating in manual drilling, and provide strong data and technical support for refined management and production of smelting enterprises.