Research On Coal Gangue Identification Technology Based On Alex Net Convolutional Neural Network

Coal is one of the important energy components in our country,and the existence of gangue in coal affects the quality of coal seriously.Traditional gangue selection technologies such as wet gangue selection and composite dry gangue selection have problems such as serious environmental pollution and rapid equipment loss.In order to reduce the pollution of coal to the environment in the process of gangue selection,image gangue selection technology has gradually matured with the development of neural networks,and has become one of the important selection methods for coal gangue identification due to its advantages of low cost and high speed.As the key technology of image recognition,convolutional neural network is of great significance to the research on coal gangue recognition.A large number of experimental results show that the convolutional neural network can realize the correct recognition of coal gangue images.However,due to the huge number of data sets required to train the parameters of the convolutional neural network,too small image data will affect the accuracy of coal gangue image recognition.And in the actual working environment,a large amount of pulverized coal will adhere to the surface of the gangue,which interferes with the correct identification of coal gangue.Therefore,in order to solve the problems of insufficient data set and low recognition accuracy in coal gangue identification process,this thesis mainly carries out the following work:(1)In this thesis,the Alex Net convolutional neural network is used to realize the correct classification of single sample images of coal gangue.In the experiment,50 coal gangue samples were collected,a total of 1500 coal gangue image data were obtained,the network parameters of Alex Net were trained 500 times,and the optimal Alex Net network was selected by analysis and comparison.For the validation set,the coal gangue recognition rate was 98.67%,which proves that the Alex Net convolutional neural network has the advantages of high recognition efficiency and less pollution in the coal gangue recognition method.(2)The coal gangue sorting technology based on convolutional neural network solves the protection problem of traditional ray gangue sorting.But the datasets required for convolutional neural network training are very difficult.This thesis proposes a new Alex Net gangue recognition technology based on texture blocks.The dataset was constructed based on the texture blocks of coal and gangue.1890 original images were collected from 77 samples,and 20,000 texture blocks were extracted.The results of the validation set show that the classification accuracy of texture blocks is 99.8% for gangue and 99.7% for coal;the overall classification accuracy of 500 coal gangue images is 97.2% for gangue and 100% for coal.Based on the Alex Net convolutional neural network,experiments with different resolutions,multiple convolutional neural networks and heterogeneous image data were tested,all of which could achieve correct classification.The proposed method for classification of coal gangue based on texture blocks provides a new idea for the classification of small sample datasets.(3)Finally,according to the influence of the coal powder attached to the coal gangue identification accuracy in the actual coal mine production environment,threedimensional data is introduced for coal gangue identification in this thesis.By simulating the production environment of coal mines,a binocular line structure light coal gangue 3D imaging system is built to collect 3D information of coal gangue;a 3D data set of coal gangue is constructed to complete the training and identification test of neural network parameters.The test results show that the recognition accuracy rate of neural network is 96.7% for the cleaned coal gangue 3D data,and 95.6% for the uncleaned coal gangue 3D data.The experimental results confirmed the effectiveness of using three-dimensional data to identify coal gangue,and eliminated the influence of adhered pulverized coal on the correct rate of coal gangue identification under field conditions,it is of great significance for the application of coal gangue identification technology in actual factories...