Ash content is one of the most important properties of coal quality and the ash prediction of coal slurry in floatation is urgent and important for automation of the floatation process. The aim of this paper is to propose a method of ash content prediction for flotation tailings by the use of image analysis. The mean gray value, energy, skewness and coal slurry concentration are highly correlated with coal slurry ash content by correlation analysis based on experiments while the particles’ size has little effect on the ash. Single variable linear prediction model between coal ash content and mean gray value was developed by the LS and its prediction errors were below 7%. For improving the prediction results, an ash prediction model based on GA-SVMR was established with additional three input parameters: energy, skewness, coal slurry concentration. This model has a higher accuracy with predictive errors all below 5% and 80% of them less than 3%. Results indicate that GA-SVMR model has a higher precision compared with LS model and PSO-SVMR model and soft-sensing model based on image features of the slurry can be used as a new method for ash detection of floatation tailings in automatic control process of coal flotation.