Feature Based Face Recognition for Twin Identification Using Images

Abstract

Face recognition system have lot of challenges due to images with different lighting newlineand background, various poses, poor acquisition of images and similarity between the newlinesame persons. Identical twin identification is a tedious task in face recognition and newlineit is highly essential in the sensitive video surveillance applications, where an entry newlineto a particular service should have been authenticated before availing the same. The newlineproposed system is focused on identifying the identical twin with three different models newlinesuch as authentication based on distance, multiple features selection and Fusion newlineScore method. In the distance method, Histogram Oriented Gradient (HOG) and Elastic newlineBunch Graph Matching (EBGM) are used to extract the information from the face newlineinput image. Twin identification is carried out based on the extracted features. The newlineGray Level Covariance Matrix and Gabor filter are used to extract the features for twin newlineidentification. In Fusion based method, the Principal Component Analysis, Local Binary newlinePattern, Histogram Oriented Gradient, and Gabor methods are used to extract the newlinefacial features and three levels of fusion such as Decision Level Fusion, Score Level newlineFusion and Feature Level Fusion are formed with the extracted features for twin identification. newlineThe Particle Swarm Optimization is used for the significant feature selection newlineand Support Vector Machine classifier is used for classification. The proposed system newlineuses ND-Twins 2009-2010 dataset of Notre-Dame University, USA which contains different newlinelighting and various poses of twins images. The fusion based method has given newlinebetter accuracy for twin identification. newline