Face Verification And Interval Calculation In Train Operation Diagram Management System

The technical data of train diagram mainly includes 25 items,such as train interval time.With the rapid development of China’s railway construction and the continuous updating of map adjustment documents,the safe and effective management of train diagram technical data is of great significance to ensure the safe operation of trains.In order to ensure the safety of train diagram technical data,it is necessary to ensure that only authorized personnel can edit and modify the technical data of train diagram.The vigorous development of deep learning makes it widely used in face recognition and verification.In this paper,face verification based on deep learning is selected as the way of authorized access,and the technical data management system of train diagram is developed.The main contents of this paper are as follows:(1)This paper summarizes the development of face recognition algorithm,focusing on the characteristics of each algorithm and the improvement ideas and direction compared with other algorithms.This paper reviews the composition and development of convolutional neural network,and summarizes the common CNN network structure and face recognition data sets in recent years.Several commonly used softmax loss functions are analyzed in detail.(2)In order to solve the problem of small optimization space caused by too strong constraint of angle softmax loss,a novel algorithm based on angle interval softmax triplet loss(AS-TL)is proposed.The triplet loss supervision is added to normalize the weight characteristics.The efficiency of feature selection is improved by adding 1×1 convolution kernel and pooling layer among different residual units.The proposed method is used to fine tune the pre-training model to alleviate the problem of difficult convergence caused by strong constraints on difficult samples.In the data preprocessing stage,the original image is flipped horizontally to enrich the facial feature information.Compared with the methods with different training conditions and model sizes,the proposed algorithm can achieve better performance only by using a single model and small training set.(3)A face verification system based on convolutional neural network is built.Based on the model of the algorithm,a real-time face verification system is developed,which can determine the identity of the face in real time.Participate in the development of train diagram technical data management system,and apply the face authentication algorithm proposed in this paper to the user login authentication,data entry authentication and data modification authentication of train diagram management system to ensure the safety of train diagram technical data.(4)Through the analysis and derivation,the constraint relationship between the factors of a kind of train tracking interval in the technical data of train diagram is obtained,and the simulation calculation of tracking interval time is carried out for various factors such as different hourly speed,acceleration,partition length,etc.