Research On Living Body Detection Method In Face Recognition System

Biometric technology has been known and accepted by more and more people,and the security of the biometric technology system is becoming more and more important.In recent years due to the lower cost of hardware and the gradually developed technology face recognition technology has been greatly developed and applied in many fields such as system login,access control system,access security,finding the criminals and so on.However the facial images or videos are easily obtained by others through some mean ways some criminals try to use legal user's face information to attack the face recognition and authentication system which seriously threats the security of the face recognition system.Although the facial live detection technology has made some progress in recent years,the security is not highin actual application,extra auxiliary equipment and additional needs of users are needed.These shortcomings greatly limit the promotion of related products.In order to provide one cheap facial live detection method without extra auxiliary equipment and with less coordination of users,this paper compares the environment features and the physiological characteristics,thoroughly studies various of facial live detection technology at present,and compares their advantages and disadvantages,then proposes a kind of simple and effective live detection method.The main work of this paper includes the following three aspects:Deeply study the main frame of current research on live detection,expound the significance of live detection technology and review current status of human facial live detection technology.Summarize and classify current anti-spoofing techniques.Describe the ideal anti-spoofing characteristics,summarize the live detection technology in the current biometric technology,and analyze the advantages and disadvantages of each method,summarize the characteristics of current live detection technology,and combine the background of this study,evaluate each technique one by one,finally propose the human face detection technology in this paper.Briefly review the face recognition system,the preprocessing technologies,face detection and classification.Expound the advantages and disadvantages of different technologies considering the time efficiency and accuracy.Finally choose the statistical method based on Adaboost to detect the image in the face and then detect the facial feature points combining the detection methods of facial feature points and feature part proposed by foreign and domestic scholars in recent years.Divide the facial detection methods into six categories according to the type of basic information on the current methods of face detection then respectively describe and analyze these methods,and finally determined to use Supvised Descent Method to extract the facial feature points,and describe the principle and process of the SDM method in detail.In view of the existing human face detection technology,in view of the existing face biopsy detection technology screening,analysis of the advantages and disadvantages of each technology,comprehensive consideration of hardware requirements and user coordination,and finally put forward the algorithm in this paper.The algorithm in this paper can be divided into four steps: firstly considering the limitations of single image human eye detection technology,this paper adopts the detection technology for video sequence;secondly,during the detection process of human eyes in the video sequences in the detection process,this paper redefines the eye opening and closing degree,adopts the average of single eye feature points and the average value of the eyes opening and closing degree also uses other methods to prevent the affection by mutations feature points;then define the standard of live detection for a single image video sequence and detect the blink of the human eye in a single image video sequence;at last,define the living detection standard of the image sequences,and finally determine the living condition of the current human being.Only after the above four stepsthose stilldetected to be living face can be finally regarded as living human,otherwise it is not living.In the end the speed and accuracy of the algorithm proposed in this paper will be tested.This algorithm is implemented on Intel(R),Core(TM),i7-3520 M,CPU@3.70 GHz,and8GB RAM hardware platforms based on Opencv2.4.10.The speed can reach 80 frames per second,and the accuracy can reach 98% in the recorded living and non-living samples.The human face detection method proposed in this paper has met the requirement of real-time application under the premise of ensuring the accuracy.