Abstract: To tackle the problems of low coal flow recognition accuracy and slow response speed in belt conveyor systems, which fail to satisfy the requirements for adaptive matching control between coal quantity and belt speed, this paper proposes a novel design scheme for a visual detection system of conveyor operating status. The system can accurately monitor the coal flow on the conveyor through visual detection and deep learning. The system utilizes a visual monitoring system to extract, segment, enhance and analyze images of coal flow, and utilizes deep learning theory to optimize data analysis and calculation schemes for coal flow detection. According to practical applications, the new control system has a detection deviation of less than 0.2% for coal flow, laying the foundation for achieving automatic adjustment of coal quantity belt speed of the conveyor and improving its operational economy.