Quantification Of Facial Information Features And Its Application In Video Analysis

Face recognition has become the most important biometric technology in identity authentication and has been widely used in many fields,for instance,military,finance,public safety,daily life and so on.In recent years,due to the rise of deep learning technology,people have changed the research pattern of face recognition.Since then,the application of deep network architecture to obtain human facial features has represented a huge advancement in facial recognition technology and has promoted practical applications successfully.Targeting at the issue of face recognition,this article perf orms the following tasks:Firstly,this paper applies Convolutional Neural Network(CNN)and adopts the End-to-End approach to learn the embedding of face images into feature vectors directly.Taking advantage of the network architecture of Goog Le Net,the loss function proposed by the Face Net model is reconstructed,and a new method of triple selection is designed to realize the extraction and quantification of facial information features.In the meantime,we compare and analyze the new model with the original model.For the multiple evaluation methods and indicators defined,the performance of the test data set shows the effectiveness and superiority of the new model.Secondly,as far as security precaution and control issues are concerned,the current security facilities are backward,the number and types are few,and the extent of refinement is not high.It is still based on video surveillance and vehicle bayonet and depends on the manual labor whose work efficiency is relatively low.With the consecutive advancement of security construction,there are progressively increasing monitoring points so that the camera generates a large amount of video data.At present,security video is generally stored for one to three months.Unstructured forms exist and are u seless except for playback.In response to the issue,this paper puts forward a new algorithm which combines face recognition technology with artificial intelligence to analyze the video content and detect abnormal information.As a result of that,public security prevention can be carried out and risk detection can be proceeded.