Research And Implementation Of The Public Security System Face Recognition Algorithm

Since been brought up in60s last century, the face recognition technology is always a heatedtopic among researchs. Especially in recent years, with the rapid development of the public securityimage network, intelligent monitoring and the public security image cloud computing technology,face recognition technology is transforming from theory to practical application in the fields ofsuspect identification, site controlling, special population identification, census registermanagement and so on. The content introduced in this paper is a subitem of Nanjing public securityintelligent image network and hybrid-network communication platform project. This paper mainlyfocuses on research and implementation of face recognition algorithms.This paper firstly introduces the the face recognition generation development process andpractical significance and discusses the main application scenarios and current research focus anddifficulty of face recognition in the public security system. Secondly, it discusses the adoption ofAdaBoost algorithm for face detection to meet the requirements of real time accuracy. Following isthe comparison of major algorithms of face recognition. This paper uses the HMM-based facerecognition algorithm which satisfies the expansibility of large face database of the public securitysystem. In the feature extraction stage, to identify the picture this paper uses2D-DCT whichresulting low frequency coefficients as the feature vector, reduces the amount of computationcomparing to the gray vector as feature, meanwhile improves the robustness of the recognitionprocess for the posture and facial expression. Finally, this paper presents the design andimplementation of a face recognition prototype system based on OpenCV. Through repeated testingpractices, gets an actual running code empirical formula to calculate normalized similarity, Thispaper uses standard face database and self-build face database to test the performance of theprototype system. The experiment results prove that the prototype system has a good performancein terms of the efficiency and effectiveness of face recognition and reaches the expected goal.