Automatic Classification System For Mine Car Loading Based On Convolutional Neural Network In Coal Mine Auxiliary Shaft

For a long time, the coal mine auxiliary shaft track transport system of China, including identification harvester loading type and controlling turnout manually, has the problems of low efficiency and high labor cost, and restricting the transportation automation of the coal mine auxiliary shaft. This paper established a classification data set of mine car loading images, proposed a method for mine car loading images based on convolutional neural network. Classifying the images of mine car loading by convolutional neural network technology solves the problems of difficult manual feature extraction and poor adaptability to environmental changes in traditional machine learning methods. Evaluated the actual performance of two typical convolutional neural networks of ResNet and SqueezeNet in the classification of mine car loading on the NVIDIA JETSON TX2 platform. The classification accuracy on the data set reached 97.6% and 98.4%, and the classification speed reached 21fps and 71fps, field test in Gucheng Coal Mine of Linyi Mining Group verified the feasibility of the method.