Abstract: In order to recognize one's identity accurately, an algorithm of feature extraction and matching based on regional shape was presented. The vein image was transformed into binary image through preprocessing first, the binary image was then thinned, and skeletal information of the vein image was obtained. Afterwards, the four connecting areas were labeled, then the endpoints and crossing points which are linked with the connecting area were found. The edge of the connecting area was approximated with several lines through the minimal distance method, and the angles of these lines were regarded as the feature of vein image. Finally, the local region was matched through the improved longest common subsequence method, and the global image was matched by the improved Hausdorff distance method. The experimental result shows that the acquired image feature by this algorithm has stronger distinguishing, the recognition result is affected less because of information missing in collecting vein images, and the correct recognition rate exceeds 96%.