Framework For Human Recognition From Multi Angular Images

Abstract

Face recognition is an authentication process that identifies face by its earlier stored database face images. The concept of face recognition is extensively studied over the past four decades for security purpose. Face recognition is still very challenging issue in surveillance system either in known environment or in uninhibited. Application of human recognition is important requirement in prevention of any unauthorized activity or trigger events followed by human based action after successful face recognition. The majority of modus operandi utilizes face recognition method produce better performance newline newlineonly when they are provided with front facing mug shot face images. Considering these challenging issues with reference to current surveillance system, the motivation of the proposed research is on side angle face images for recognition. newlineThis thesis proposes framework of human recognition using feature based method. Feature based method is works upon individual face features rather than harmonising whole face template. The main objective of human recognition framework is identification of human having front and side view face with higher recognition efficiency. This research work mainly emphasizes on i) Human Face Detection from Single person image or Group photo, ii) Cross Face Orientation, iii) Face Feature Extraction, iv) Human Recognition by Face Feature Matching. Face Detection algorithm accumulates only face(s) region. newline newlineThe detected face is examined and if found cross, then it is processed into straight face using face rotation in face alignment component. Facial feature components like eyes, nose and mouth are extracted using basic human face template and combination of edge detection and corner detector. Feature comparison performed by comparing feature boundary s coalition in which calculation of matched points is measured. The Euclidian distance method is used for feature matching. The maximum match value is retrieved among all database images, the proposed solution decides whether human face... newline newline