To solve the problem of single and incomplete feature extraction types in ash detection based on flotation tailings images, a method for predicting the ash content of tailings based on machine vision multi feature fusion is proposed. An industrial tailings image dataset was obtained on the flotation field. RGB (Red, Green, and Blue) color features, grayscale features, gray level co-occurrence matrix features and color co-occurrence matrix features were used to describe tailings images. The relationship between features and ash content of tailing was studied through corre-lation matrix. Principal component analysis (PCA) was used to reduce the dimensions of primary feature space, principal components with different numbers were used as inputs, tailings ash content was used as output, and a support vector regression (SVR) model was built to predict the tailings ash content. The experimental results show that the multi feature fusion significantly improves the accuracy of the tailings ash content prediction model, provides a more comprehensive description of the tailings characteristics, and performs better than models that use a single type of feature as input. This method can provide a theoretical basis for the intelligent construction of flotation.