Research On Identity Verification And Anti-tailing Alarm System Based On Machine Vision

With the rapid development of machine vision,biometrics technology is gradually integrated into people's lives.As one of the innate biological characteristics of human body,human face is strongly distinguishable among different individuals,which provides a prerequisite for identity verification and authentication.At present,the identity verification based on face recognition technology has been widely studied and applied in the field of transportation,providing important guarantee and support for traffic safety.The identity verification system includes three necessary steps:face detection,feature extraction and comparison.However,in practical applications,the identity verification system is highly susceptible to environmental factors.Especially in traffic security scenarios where the background environment is complex and variable,queuing and congestion are likely to cause inaccurate face detection,which affects the performance of the identity verification system to a certain extent.At the same time,the existing system has a low intelligent level in distinguishing trailing events of pedestrian,which affects the further improvement of security check efficiency.Aiming at the problem of inaccurate face detection and insufficient intelligent level in the existing identity verification system,this paper designs an identity verification and trailing discriminant algorithm based on convolutional neural network(CNN).First,the facial features are extracted according to the Haar feature template and the face detector is trained based on Adaboost.Then,a network model of multi-scale prediction is constructed,and a pedestrian head and shoulder detector is trained.After that,the face information with head and shoulder information are matched in the detection results of the two cameras mounted at different positions According to the layout relationship,combining target scale and positional relationship,the "first face" is determined in the complex background.After that,a convolutional neural network of multi-layer feature fusion is constructed by optimizing the model parameters.The optimal face recognition model is obtained,and the judgment result of "first face" is used to compare the human to the photo of ID card.Finally,the trailing event is discriminated based on the target number and position relationship in the head and shoulder detection results in the channel area after one time verification.On the basis of the algorithm implementation,this paper designs and develops a set of identity verification and anti-tailing alarm system,which controls the status of the gate according to the comparison result of human and the photo of ID card,and alarms the trailing event.In the simulation scenario test experiment,the system can run stably around the clock,the pass rate reaches 96.50%,the rate of trailing event discriminant is 79.50%,and the average elapsed time from card reading to gate opening is 2.25 seconds,which verifies the good performance of the system under complex scenes.