The Real-Time Face Recognition Based On Embedded System Of Low-power

In recent years, with the development of face recognition technology,face recognition system has been applied to engineering practice fromresearch works. It has been used for video surveillance, attendance&access machine, Immigration Auto-Clearance System and so on. There aresome advantages of face recognition that non-contact, uneasy tocamouflage, low acquisition equipment, compared to irides, voicerecognition system and others. But there are some disadvantages of facerecognition based on PC like large size, cost high, high power and uneasyto carry, and it restricted the application. Embedded ARM processor is lowpower consumption, powerful data processing ability and low cost. It isbeneficial to universalness of face recognition. Hence, the face recognitionsystem based on arm will have a better development prospect forapplication.Aiming at developing real-time face recognition based on embeddedsystem of low-power, this thesis proposes designing and implenment thesoftware and hardware based on Embedded ARM processor, researchsand improve the face detection and recognition algorithms, accomplishsthe development and optimization of face recognition algorithms andapplication software coat, in final, accomplishs the total system.The major tasks in this paper could be represented as follows:The software and hardware architecture of embedded facerecognition system are designed based on ARM platform,acocording to the requirement of face recognition algorithm onembedded system. The face recognition algorithm and application software coat is separated layer. The functions ofmulti-interface and low power are implemented.The eye localization algorithm is refined in this thesis. Themethods of eye localization algorithm based on eye search andselecting are proposed to solve the disturb of eyebrow, based onthese, for ensure the quality of face pictures, the algorithm ofeye localization accurate is investigated to get rid of the false eyelocation pictures and the algorithms of this paper is improved.In order to solve the problem of light conversion and limitedspeed, the LBPV algorithm is proposed to use in face recognitionfor features selected in this thesis. LBPV is an improved algorithmof LBP and can describe both spatial and asymmetry feature oftexture, also it improved the recognition rate.The face recognition algorithm is optimized on program level onbased of algorithm. The speed is increased a lot, with little effecton recognition performance and enhancement on real-timeproperty.Application interface layer and flow charts are designed andimplemented. The module of algorithm is package andapplication interface layer can call algorithm function. In final,face recognition based on embedded system is accomplished.