Face Recognition Based On Near-Infrared Images

As the development of society and technology,more and more verification is needed,such as access control,ATM system of bank,and account login in internet interface。Compared with traditional verification ways,face recognition is of more security and convenience,and could void troubles like missing passwords and keys。Compared with other human biological features,such as iris and finger,face recognition is more friendly and easy to capture。There is no doubt that face recognition technology will change the life of human beings greatly。Face recognition technology has made great progress within tens of years,a lot of algorithms works well。But practical face recognition system still face with many disturbances which hasn't been solved,such as pose and illumination changes。To make system more robust to illumination changes,we employed near-infrared images。The main work of this thesis can be described as follows:1. We designed and carried out the plan and process of capturing near-infrared face images。The pictures is very important for training and test face recognition algorithms。Unfortunately,almost of all face database is captured under visible light。Then we collectednear-infrared face images according to those capturing plans。2. Eye localization on near-infrared images。Face detection and localization is one of the most important task in face recognition systems,but ways on visible images doesn't work as well on near-infrared images。We obtained reliable localization results by training with near-infrared images。3. Processing on near-infrared images。What we did can reduce image disturbance,and enhanced image features。What's more,we compared several image processing method on near-infrared images。4. We proposed some improvements to speed up LBP processing。 5. Facial feature selection。It's proved that different face areas have different importance in face recognition。We obtained better result by selecting most discriminate features using AdaBoost。...