Abstract: It can effectively save energy and reduce consumption for mining enterprises to adjust the speed by frequency conversion by monitoring the coal quantity dynamically . .In view of the existed hardware resources and communication interface of mine, here a dynamic identification and optimization scheme based on image processing is proposed: An adaptive filtering method based on window threshold is proposed to solve the problem of image reflection. It removes the noise effectively while retaining boundary information. In the stage of image segmentation and extraction, a modified adaptive threshold segmentation algorithm based on classification method of EMD is proposed to solve the problems of low boundary contrast and unstable field light sour. The image is processed by the method of first etching and then expanding to improve the boundary recognition accuracy. A continuous frame difference method is proposed to identify the running state of belt. The results of site test show that the system has the advantages of high-performance, which requires no additional equipment. It deserves the value of popularization and application.