A Coal Gangue Identification Method Based on HOG Combined with LBP Features and Improved Support Vector Machine

Identification of coal and gangue is one the important problems in industry. To improve accuracy identification mining process, a method based on histogram oriented gradient (HOG) combined with local binary pattern (LBP) features improved support vector machine (SVM) was proposed. First, according to actual underground working environment mine, vision platform for built image acquisition experiment carried out. Then, images were denoised by median filtering, extracted using HOG LBP feature extraction algorithm, these normalized principal component analysis (PCA) reduced dimension remove correlation redundancy between features. Finally, SVM, SVM optimized genetic algorithm (GA-SVM), particle swarm optimization (PSO-SVM) grey wolf (GWO-SVM) used as classifiers classification, respectively. The experimental results show that GWO-SVM classification model has highest accuracy, average accuracies 96.49% 94.82% training set test set, respectively, which shows optimize good effect images, proves feasibility proposed method.