Segmentation and registration on near-infrared hand vein images for injection and personal identification

To find the veins in intravenous emergency injections, non-invasive feasible methods seem to be necessary. In another area in biometric identification, hand vein patterns can be used as other external identifiers are more probable to be damaged or forged. To achieve these goals, NIR back hand vein images are captured. In the proposed method, the NIR hand vein images are pre-processed and segmented by means of Maximum curvature algorithm and hand veins are extracted. The segmentation accuracy is significantly good for the purpose of emergency injection. For the identification purpose, after segmenting each image, the resulting image is correlated with the segmented images in the database and the unknown input image is assigned to the related image in the database. The identification precision is 100% without any hand displacement as the hand position is a predefined location. To simulate the unpredictable hand position displacement, the images are artificially rotated and robustness of the identification procedure to the rotation is 12° in the worst case.. And by using registration, the robustness of the identification procedure is 98.6% with any degree of rotation.