LAUSR

Abstract At present, a lot of features have been extracted to characterize the froth flotation, but there exist redundant and irrelevant features which may degrade the performance of a classifier and influence the production condition recognition. In this study, a feature selection strategy based on the minimal-redundancy-maximal-relevance criterion (mRMR) is used to find the most useful but less redundant features. Additionally, the least squares support vector machine (LSSVM) optimized by state transition algorithm is proposed to serve as the classifier in feature selection. It is found that hue and energy of high frequency play significant roles in classification of flotation froth images. Experimental results show the effectiveness of the proposed method.