Research on coal gangue classification recognition method based on the combination of CNN and SVM

For the traditional machine learning methods rely on manual experience and deep learning classification model depth and complex structure, resulting in poor gangue classification performance, this paper proposes a coal gangue recognition method (CNN-SVM) based on the combination of convolutional neural network (CNN) and support vector machine (SVM). Firstly, we use a generative adversarial network (DCGAN) to generate new coal gangue samples and expand the gangue dataset by traditional image enhancement techniques to increase the data samples and improve the generalization of the model; then we construct an efficient and simple CNN as a coal gangue feature extractor and verify the effect of convolutional kernel size on the accuracy of the model, and determine the 5×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document}5 size of the convolutional kernel to extract more accurate and comprehensive coal gangue; Finally, it is combined with SVM using grid optimization to improve the accuracy of coal gangue recognition. The experimental results show that the recognition accuracy of the constructed model reaches 97.5%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document}, which has obvious advantages compared with traditional classification models and classical classification models, and the recognition speed is faster compared with the mainstream classification models, which provides a new idea for coal gangue recognition.