Research On Face Detection And Recognition Algorithms Based On Machine Vision And Implementation In Escalator Scene

With the rise of artificial intelligence technology,intelligent and convenient life has gradually become the pursuit of people,and the development of convolutional neural networks has made the image field face an unprecedented opportunity for development.Therefore,computer vision technology has been applied to all aspects of real life,and face recognition is one of the fastest growing technologies.Face detection algorithm,data processing algorithm and face recognition algorithm in the escalator scene were studied,and a face recognition system applied to the escalator scene was designed and implemented.The image was acquired by the camera installed in front of the exit of the escalator,and the face of the passenger in the image was recognized,thereby realizing functions,such as visitor identification,employee identification and passenger information acquisition.The main work of this article includes four aspects:1)Aiming at the problem of unstable performance of face detection algorithm in complex scenes,a thorough analysis of various methods was made.The data set for face detection suitable for the escalator scene was firstly established.After that,three features of Haar,LBP and HOG and four detection algorithms of SVM,Adaboost,cascaded CNN and MTCNN were studied.MTCNN was selected as a face detection algorithm for escalator scenes through comparative experiments on data set.2)Aiming at the problem of regional specificity of face recognition algorithm,the actual scene data was obtained.Firstly,two data sets,suitable for face recognition in escalator scenes,were constructed.After that,the traditional data augmentation algorithm was used to solve the problem of less data.The GAN algorithm proposed in this paper was used to solve the category imbalance problem.3)Aiming at the problem of matching the existing data set with the model capacity,a network model based on improved residual network and center loss was proposed in this paper.Firstly,the basic model of the network was determined according to the amount of data and the complexity of data.The loss function of the network was the combination of center loss and cross-entropy loss.Then some of strategies were used to optimize the model,such as BN,dropout and regularization techniques.Finally,the results of longitudinal contrast experiments and lateral contrast experiments show that the model has good ability to extract features.4)Based on the above research results,a face detection and recognition system in escalator scenes was developed by using Python,OpenCV,database and other tools.The software includes initialization module,information transmission module,function module and escalator monitoring module.The system functions include registration,login,employee identification,visitor identification and so on.We focuses on the theme of "face detection and recognition algorithm in the scene of escalator",which mainly includes three parts: face detection algorithm,data set processing algorithm and face recognition algorithm.The experimental results show that the proposed algorithm achieves a high accuracy and the system is of great significance to realize the intelligentization of escalator.