Abstract: A multi-pose hand vein recognition algorithm was proposed based on 3D point cloud matching. Considering the characteristics of hand vein point cloud,an extended database that combines 3D feature arrays and vein point cloud data was established according to stereo vision principle. A 3D feature array based calculation method was proposed for coarse point cloud registration. In stereo vein image, the stable feature points were extracted and reconstructed to three-dimensional features. The posture difference of hand vein point clouds was eliminated according to the result of 3D feature matching. An improved normal distribution transform algorithm was used to complete the vein point cloud matching. Experiment results show that the proposed algorithm can effectively improve the recognition rate under multi-pose. The recognition rate of the system can be more than 90%, even if the hand posture changes in a large range.