Research On Fast Face Recognition Algorithm Based On Binocular Vision

The main use of biometric technology with the body's own characteristics,including physical characteristics and behavioral characteristics to carry on identity authentication and identification,and face recognition is a non-contact biometric technology with higher natural,acceptable and uniqueness which make face recognition technology one of the hot spots in biometrics.This dissertation introduces the background and significance of face recognition,and summarizes the research status of face recognition.This dissertation introduces the basic knowledge of binocular vision theory,and expounds the geometrical model and camera model of binocular vision,and introduces the technology of face recognition.In the 3D face recognition,the complex algorithm and too much computation make the recognition speed slow,but the 2D face recognition rate is relatively low.In this dissertation,a fast face recognition algorithm based on binocular vision is proposed,which combines 2D face recognition technology and 3D information.Firstly,the face rectangle region is extracted by feature extraction,and the facial area is obtained by facial pixel statistic;Then,according to the theoretical knowledge of binocular vision and the relationship between the map area and the actual area,the face rectangle area of the face is calculated;Finally,according to the value of face rectangle area to construct face sub-database.A stereo vision platform based on binocular camera is built with the application background of indoor intelligent service robot,The face images of 91 college students were collected,and the binocular database was established.The experimental verification of the proposed algorithm was carried out using this database.When doing face recognition experiment,the principal component analysis algorithm is used.There are two situations:considering the measurement error,the highest increase of face recognition rate is 3.27%and the maximum increase of face recognition speed is 26.12%comparing to the situation when the sub-database is not constructed,when the measurement error is not considered,the face database is divided into five sub-database,the increase of face recognition rate is 8.75% (?) the increase of face recognition speed is 51.1%.