Research On Key Problems Of Face Recognition System In Coal Mine Based On Deep Learning

Face recognition is a biometric technology for identity recognition based on human face feature information.Compared with other biometric technologies,it has the advantages of simple use,zero contact and protecting personal privacy.So far,face recognition technology in general scenes has been very mature.However,face recognition in special scenes and environments still faces many problems.In this paper,the performance of face detection and face recognition is greatly reduced due to the influence of illumination,face ash occlusion,pose change and face angle.These key problems in face recognition are deeply studied.In order to meet the requirements of network training,this paper uses the method of geometric transformation to expand the basic sample set,At the same time,through the comparative experiment,Gaussian filter and limited contrast adaptive histogram equalization image enhancement method suitable for coal miners’ face image are selected to preprocess the sample image,so that the pixels of the sample image are evenly distributed and the facial contour feature information is more obvious;By comparing the general face detection algorithms,the face detection based on active shape model is studied.The model simulates various situations of face information in the image through geometric transformation,trains local deformation to establish the face shape model.The experimental results show that the model can be well suitable for workers’ face detection in special working environment in coal mine;By adding bottleneck hop structure to optimize RESNET network structure,BN layer and dropout layer are added to solve the problem of over fitting when the number of network layers is deepened.CBAM attention mechanism is added to the feature extraction part of RESNET network structure,and the spatial relationship of feature map is used to strengthen the texture feature information weight of key facial regions and capture more facial contour features.Based on the extended face data set,this paper tests and analyzes the optimized network model.The test results show that the optimized RESNET model improves the recognition speed and recognition accuracy for the special face recognition of coal mine workers.The network structure model can be effectively applied to the face recognition of coal mine workers,realizes the design requirements,and has certain practical significance.