Research On Classification Of Video Scene In Coal Mine Based On AlexNet And LSTM

The characteristics of China's energy structure determine that coal is the main energy source,and the safe production of coal has always been a major issue emphasized by the state.The application of intelligent monitoring technology in coal mine production is one of the effective ways to prevent and mitigate coal mine accidents.Scene classification of surveillance video data can not only provide valuable reference information for video surveillance work,but also improve the accuracy of surveillance results,which can provide a basis for the next investigation of abnormal situations in the scene.In this paper,mine video data is taken as the research object,and the problem of scene video image classification with complex background is studied.The main work and innovation are as follows:(1)Due to the noise,poor contrast and uneven illumination of the mine image,the quality of mine video image is poor,which seriously affects the accuracy of the mine scene classification.Therefore,a coal mine underground video image enhancement algorithm based on the Restricted feedback function is proposed.Firstly,the algorithm combines the weighted average method and the median filter to perform grayscale and denoising preprocessing on the image.The experimental results show that the algorithm can effectively improve the video image distortion of coal mine underground,make the image layer rich and clear,and has a faster processing speed.(2)In order to classify the scenes of mine video accurately,an algorithm of coal mine underground video scene classification based on AlexNet and LSTM is designed.Firstly,the input data is preprocessed,and the mine video background is extracted based on the background subtraction method.Fixed background method is used to establish the background model,and then background subtraction method is used to get the difference image.Then the difference image is thresholded by the method of maximum inter-class variance.Finally,based on the background subtraction method,the background information from the video is obtained by twice difference.Then,taking the background information in the mine video as experimental data,a seven-layer convolutional neural network is designed based on AlexNet network to extract the scene structure features.Then,a two-layer stacked LSTM network is used to obtain the temporal dynamic features in the video.Finally,the classification is carried out by Softmax classifier.Experiments show that the proposed algorithm model has better accuracy than traditional algorithms in the classification of mine scenes with complex background.