Research on Coal Gangue Identification by Using Convolutional Neural Network

Abstract

A237 at the problems of traditional coal gangue image recognition methods such as difficult extraction of artificial features and low accuracy of recognition, a method of automatic identification of coal gangue images by convolution neural network is proposed. Based on the classic convolution neural network LeNet-5, this method is improved from input sample size, activation function, network depth, size and number of convolution kernels, classification function and so on. The learning curve of the improved network shows its structure is reasonable. The recognition rate of the original LeNet-5 is only 50.67%. After the improvement, the recognition rate of coal gangue reaches 95.88%, which is much higher than that of the traditional recognition method. The results show that convolution neural network can effectively learn and extract the image features of coal and gangue automatically, classify coal and gangue, and provide reference for identification and classification of coal and gangue.