Extraction of unrestricted palm patterns in a 3D environment and its application in biometrics

With the increase of the studies on palmprint recognition for the last 10 years, they are faced with the problems that arise from unrestricted pose (different angles and distances) of the biometrics. In this study, on the purpose of increase recognition success on palmprint recognition systems, a new approach based on geometric correction of palm region's pose is presented and system success is analyzed using current approaches. Hand's 3D pose is determined using depth information which is based on detection of matching points and camera calibration and image is transformed 2D plane with geometric image transformations. Unrestricted palm pattern selection is performed with Active Appearance Model (AAM) on images transformed 2D plane. In this study, 86% recognition performance is achieved on unrestricted but uncorrected patterns. With the proposed method, the recognition performance on palmprint patterns, which are calculated from the same palmprint images by 3D pose estimation and correction, has risen to 93%.