Abstract:With the rapid consumption of phosphate rock resources, there are problems of large proportion and low utilization rate of low-grade phosphate rock in China at this stage, which leads to the increase of production and treatment costs. Instead of manual observation, machine vision technology is used to realize early discarding and preconcentration of phosphate rock, so as to improve the comprehensive utilization efficiency of ore and reduce the economic cost. According to the characteristics of Bagu medium and low-grade thin bedded phosphate rock in Leibo County, Sichuan Province, a dynamic real-time pre separation algorithm of phosphate rock based on HSV color model, multi threshold method and KNN algorithm is proposed in this paper. The concentrate grade of the ore to be sorted and tested by this algorithm can reach 18.3%, which shows that the recognition accuracy of the algorithm proposed in this paper is high, which basically meets the separation needs of enterprises to identify tailings, and achieves the purpose of tailing.