A 3D interpolation method in repairing hand vascular tree for vein recognition

Abstract

Hand vein recognition is one of the most widely used methods for authentication. However, since the restriction of fixed hand posture and the loss of vein's depth message in traditional 2D vein recognition will cause a higher false rejection rate (FRR), many researchers have focused on the 3D hand vein recognition. For vein recognition, the knowledge of the vascular tree's topology and symbolic description of its feature points will be essential, while by using the common 3D reconstruction methods the vein data obtained is sparse and not enough to describe the topology of hand vein. In this paper, a computational framework to develop vascular tree interpolation with minimum human intervention to repair the possible broken vessels is proposed, which comprises three main parts including labeling point cloud into different clusters, finding pairs of interpolation points and the process of interpolation. This achieves good performance on 18 sets of hand vein point clouds.