Face Recognition Based On Face Recognition

In order to facilitate the identity recognition of the entrance and exit people,and to reduce the security personnels' pressures on duty,the predecessors have explored the field of video surveillance on the import and export characters identification system,and they have got lots of achievements.However,there are still some problems that are worthy of further study,such as: how to reuse other positioning algorithms on the basis of face localization algorithm so as to ensure the accuracy of positioning;how to keep a good balance between accuracy and speed on the face recognition;how to extracte image information automatically recessive.Adaptive Boosting Algorithm(Adaboost algorithm)is a commonly used face localization algorithm.Convolutional neural network(CNN)is one of the key technologies of deep learning.Based on the previous work,this paper uses Adaboost algorithm and CNN to develop the entrance and exit characters recognition system.The main function of this system is used for entrance and exit characters according to real-time location,extracte image information,characters recognition so as to know who they are.When someone goes by,the system will automatically alert the attention of security personnel.The development of this paper is divided into three aspects:1.Using Gauss mixture model to extract the moving object region of interest,the detection range is reduced effectively for the next step of face localization.2.Propose a method of face localization based on Adaboost algorithm and CNN.Firstly,the Adaboost algorithm is used to find the face region of interest in the target region.Then,the face image into the CNN region again discrimination,which can improve the face positioning accuracy on the basis of further the Adaboost algorithm get the positioning results.3.Propose a method using the step 2 method of calculating the convolution layer of CNN of calculating the convolution layer of CNN.the network sampling layer is reduced in order to cut down the Face recognition time by damaging some recognition accuracies.The above-mentioned 3-point integrated system is applied to the entrance and exit scene of video surveillance and the identity recognition of the entrances and exits.The test work mainly from three aspects to test: the face localization,face recognition,the system scene.1.The face localization experiment in database of California Institute of Technology established by Wu Enda and Guiyang field scene shows that the Adaboost algorithm and CNN method are used to locate the face accuracy higher than that of Adaboost method alone.2.In the face recognition open database Yale and ORL test,the experimental results show that the improved CNN algorithm has a certain recognition accuracy,but the recognition time is reduced by 0.013 seconds per image.3.This tests work on the Guizhou Minzu University administration building,laboratory building,dormitory,kindergarten entrance.This system is used to train the original image directly in the CNN training process,and the image information feature is extracted and classified automatically.The Adaboost algorithm and the CNN method are used to effectively remove the pseudo-face image in the face location.for the next step to create the conditions for the work;and face recognition using improved CNN algorithm,at sacrifices some recognition accuracy under the premise of reducing the identification time for real-time identification to lay the foundation.The system can reduce the working pressures of the security personnel and have a certain Application value and promotion value.