FEPNet: A Feature Extraction-prediction Network for Coal Flotation Froth Image Segmentation

Coal flotation represents a flotation technique used to extract valuable mineral components from a coal slurry. The visual information on the coal slurry flotation froth's surface can both provide intuitive feedback on the flotation effect and guide the operator to adjust the flotation process parameters accordingly. However, the image segmentation of the coal flotation froth is challenging due to the problems of high noise, fuzzy boundaries, and strong adhesion. To address these problems, this paper proposes an innovative and effective segmentation model for coal flotation froth images named the Feature Extraction-Prediction Network (FEPNet). The FEPNet is a deep neural network model with an encoder-decoder structure, and it mainly consists of two parts, namely a feature extraction network and a prediction network.The network model is designed using the Stem Convolution Group module (SCG), Global-Local Hybrid Attention module (GLHA), and High-Width Combined Attention module (HWCA) to extract global and local feature information and realize multi-dimensional feature information fusion. The proposed FEPNet model is evaluated on a self-constructed image segmentation dataset of coal slurry flotation foams. Compared to the UNet model, the proposed model can improve the mIoU, DSC, and Acc metrics by 4.86, 2.86, and 3.48%, respectively. The experimental results show that the proposed FEPNet model can effectively address the segmentation difficulties of coal slurry flotation froth images. This study lays the foundation for optimizing the flotation process parameters and realizing the intelligence of flotation field parameters. In addition, generalization experiments are conducted on the skin lesion segmentation dataset, and the results show that the proposed FEPNet has excellent segmentation generalization ability.This paper provides a basis and reference for the automatic control of flotation process, and lays a foundation for improving production efficiency and product quality.