Research On Face Recognition Of Invaders In Natural Gas Gathering Station Based On Deep Learning

The unattended station makes full use of advanced technologies such as Internet of Things and automation to realize the all-day autonomous management of the station.This method not only effectively reduces the consumption of manpower and material resources,but also significantly improves the management level of the station,which is one of the development trends of intelligent construction of natural gas gathering stations.However,the station is generally located in a remote area.Once the suspicious personnel invade,it is difficult for the unmanned management to monitor effectively in the first time.Therefore,higher requirements are put forward for the safety protection technology of the station.In this paper,the face recognition technology is applied to the security management of unattended stations.The deep learning method is mainly used to solve the problem of face recognition under the condition of single sample in order to improve the security management level of unattended stations.The main research contents are as follows:(1)A multi-feature face generation model based on improved Generative Adversarial Networks is proposed.The unmanned station is in an open environment,and the limitation of a single face sample greatly affects the effect of face recognition.To solve this problem,we carry out the research of multi-feature face generation model based on Generative Adversarial Networks.The network is trained by using CelebA dataset containing rich feature attributes.The face samples with different feature attributes are obtained by adjusting the network structure and parameters continuously,which provides effective training sample support for the next task of face recognition.(2)A face feature recognition method based on improved convolutional neural network is proposed.In order to meet the training needs of the deep network model,we continue to enhance these samples,such as random clipping,brightness transformation,translation,rotation,flip and so on.At the same time,A face recognition model based on convolution neural network composed of eight layers network is constructed.The hidden features of face are extracted automatically by deep neural network.The optimal network structure and parameters are obtained by optimizing the horizontal and vertical design,initialization parameters and optimizer types of the model.Finally,the training model is tested by the open data set,and the experiment shows that it can achieve a better recognition effect.(4)A prototype system for intrusion prevention of unattended stations is developed.The prototype system takes unmanned station as application scenario,and combines singlesample expansion,face detection,image preprocessing,face recognition and other modules to form a set of feasible solutions.Through testing in different scenarios,the prototype system can achieve the desired recognition effect.