Abstract: In order to further improve the automation and intelligence level of open-pit mining operations, reduce the labor intensity of self-moving crusher drivers, and improve the working environment, an image recognition based automatic speed regulation and block recognition method for the crusher plate feeding device is proposed. This method collects, processes, and stores real-time panoramic images of the crusher's feeding hopper. After filtering, convolutional neural network algorithms are used to extract image detail features. The improved YOLOv5 algorithm framework is used to calculate the coal quantity and category characteristics of the feeding hopper. Finally, the minimum distance classification algorithm is used to obtain the coal quantity category of the feeding hopper, and to determine whether there are large coal blocks in the feeding hopper to avoid blocking the plate feeding device. Using the above method, a visual detection system for coal quantity and large blocks in the receiving hopper is constructed, and the real-time coal quantity and large block results of the receiving hopper are sent to the PLC of the crusher. The PLC adjusts the speed of the plate feeding device to prevent situations where the feeding hopper is empty or overflowing, as well as large block blockage, which affects production efficiency.