Research And Development Of Touchless Finger And Palm Recognition System

The security system based on palmprint and fingerprint recognition has a large market share.Specific to the drawbacks of low recognition accuracy or poor anti-spoofing ability of traditional single biometrical feature recognition systems,this thesis proposes robust biometric recognition systems based on multiple biological features.More specifically:(1)This thesis presents a recognition scheme based on Distal-Interphalangeal-Crease(DIP)and palmprint,and then develops an APP for human authentication on Android mobile phone.In this thesis,a database containing 100 individuals and 400 palm images(including fingerprint and palmprint)is collected by phone camera.Experiments are carried on the established database to evaluate the performance of the proposed method.The proposed recognition method consists of four steps: 1)extracting the region of interest of palmprint using skin color detection and segmentation algorithm based on maximum connected region;2)locating the coordinates of key points such as valleys between fingers using Harris corner detection and locating the key points of DIP by the connected region and projection method;3)using the competition coding to represent the palmprint and DIP features,and a hierarchical fusion strategy is proposed to fuse DIP and palm print features;4)adopting a down-sampling method to handle data so as to ensure the system have real-time process ability.(2)Since finger vein can be used for spoof detection to prevent malicious attacks,a reliable verification system based on multi-view fingerprint and finger vein features is developed in this thesis.It mainly includes: 1)designing a touchless multi-view fingerprint and finger vein acquisition device to collect fingerprints and finger veins.Visible light is used to image fingerprint and near infrared light is used to image finger veins;2)establishing a multiview fingerprint and finger vein database containing 135 individuals and 12960 finger images(including 6480 fingerprints and 6480 finger veins),and proposing a method based on DIP and finger veins.The performance of the proposed algorithm is verified on the database;3)realizing an end-to-end application system based on fingerprint and finger vein features.Finally,this thesis further explores the three-dimensional fingerprint reconstruction technology by using multi-view fingerprint images.The multi-view images of a same finger are matched by the minutiae feature and the SIFT feature,and then the 3D coordinates of the 2D multi-view images are established by camera calibration.Finally,the 3D shape of a finger is reconstructed by the 3D coordinates and a prior estimated finger shape model.