Abstract: The coal content in gangue is one of the key indicators of coal preparation production, which cannot be detected online at present. A new method based on machine vision and particle swarm optimization-support vector machine (PSO-SVM) was proposed to detect the coal content in gangue. First, the sample pictures on the gangue belt are segmented to identify the areas belonging to the category of coal and gangue. The size feature and density feature are extracted from each region. The characteristic parameters of coal and gangue are obtained respectively, and then the feature of the gangue are divided by that of coal to obtain the two-dimensional feature of the sample picture. At the same time, the binocular images are collected on the belt, the height information of the image is obtained, the height ratio 3D image characteristics are calculated. The eight optimal features are screened by calculating the Pearson correlation coefficient, and finally the support vector machine is used to predict the coal content in gangue and optimize it through particle swarm optimization. Establish a plane feature model and a 3D feature model, and compare and analyze the prediction results of the two models. The performance of the 3D feature model is significantly better than that of the plane feature model. The average relative error is 7.57%, which was satisfactory.