A Rotation Invariant Algorithm for Bimodal Hand Vein Recognition System
Rosalina R. Estacio, Noel B. Linsangan · 2018
The use of biometrics provides a more reliable security layer in identification and user authentication in many fields of commercial and institutional transactions. Vascular vein patterns from the hand have recently been explored as another approach to the biometric modality because these innate patterns are constant, remains distinct lifelong; relatively stable, cannot be forged, tampered nor copied. This study proposes the use of a rotation invariant algorithm like ORB for a bimodal approach in the vein recognition from the palm and dorsal vein patterns of the hand. Using near-infrared LEDs, Raspberry Pi with NOIR camera module, a portable, real time hand vein pattern recognition device was developed. Image pre-processing, feature extraction, feature matching, database of user information and GUI are implemented in Raspberry Pi, Python and OpenCV Libraries. ORB was used for generating feature descriptors and BruteForce Matcher for feature matching. Match scores generated by the classifier from dorsal and palm vein were combined using sum-rule in score level fusion to generate the final recognition results. After experimental tests conducted, system performance resulted to 95.00% accuracy level and overall response time of 2. 76secs. The developed architecture can be integrated with other systems like attendance monitoring, access control, identity authentication for financial transactions, forensic investigation, and fraud detection.