Vein point cloud registration algorithm for multi-pose hand vein authentication

Abstract

Traditional vein recognition technology is usually based on the 2D infrared images of hand vein. However, the hand vein patterns extracted from 2D images are seriously distorted with hand posture changes. A point cloud matching vein recognition system based on Kernel Correlation(KC) algorithm is designed, which provides a new perspective to hand vein recognition but also suffers from posture change because KC value is sensitive to coordinate shift. In this paper, a vein point cloud registration algorithm is proposed to improve the vein authentication under multi-pose. Experiments show that the point cloud registration algorithm can effectively improve the recognition rate. After the point cloud registration, when the hand posture changes limited to the range ± 20 degrees, the system can keep more than 90% recognition rate, the corresponding error rate is only 4%.