Soft measurement method for froth layer thickness based on multi visual features

In flotation process, the froth layer thickness has a direct effect on the mineral recovery and concentrate grades. While the visual froth features have a distinctive implicit relationship with the thickness of the froth layer. A novel soft measurement method for the froth layer thickness based on fusing different froth visual features is presented. A novel color complex network-based method is proposed for extracting the froth image textures. Combined with other froth visual features, which constitute the feature vector of the froth layer thickness, the kernel principal component analysis (KPCA) method is used to reduce the feature vector dimensionality. Then, a soft measurement model for the froth layer thickness is established using the regularized extreme learning machine (RELM), with an input that is the principal component of the visual features and an output that is the predicted froth layer thickness. This method provides a new means for measuring the froth layer thickness, which leads to a more stable flotation process and to better economic performance of flotation process.