Abstract: In order to realize the effective quantitative analysis of the color characteristics of hematite reverse flotation foam and establish the correlation model with flotation grade，the analysis of foam color characteristics based on multi-color space collaborative analysis and machine learning is carried out. By an⁃ alyzing the color characteristics of hematite reverse flotation froth，65 kinds of initial color features includ⁃ ing color moment，standard deviation and relative red value are extracted，and the key color features signifi⁃ cantly related to flotation grade are selected by correlation coefficient analysis method，including the mean value of G channel in RGB color space，the mean value of Y channel and the skewness of V channel in YUV color space，and the relative red value in gray space. On this basis，the random forest algorithm is used to construct the flotation grade prediction model and the verification test is carried out. The results of mean square error of 0.47，determination coefficient of 0.73 and average absolute error of 0.51 are obtained.The key color features extracted can effectively characterize the color of hematite reverse flotation froth，and there is a strong correlation between the color and flotation grade.It can be used as an effective tool for visual identification of hematite reverse flotation conditions，which is of great significance for the development of hematite flotation condition monitoring system based on foam color change and the improvement of flotation automation and fine control level.

Key words: hematite reverse flotation, image recognition, color characteristics, foam state quantifica? tion