| Vein pattern recognition is a kind of infrared biometrics. It has not only high recognition ratio and stability but the friendly acquisition process and the easy operating. In addition, it will not left out personal information for the veins located in internal body and very strong universality and uniqueness causing the process without contact of vein image acquisition. With these advantages, vein recognition is both paid more attention than many other biometric technology, such as face, ears, retina, iris, speech, the shape of the hand, palm print, fingerprint, signature, gait, and so on, and attracted numerous researchers to study and improve this technology.Hand vein recognition just gets preliminary results now. It can be proved by its usage convenient, comfortable, safe, reliable, strong immunity to forgery, etc. However, alike some of the popular biometrics, hand vein recognition still exist some problems for a business system. As there are still some problems such as so much cost time in consumer system implementation, so low of recognition rate, so complex really environment of acquisition vein images cuasing hard to recognise etc, need to be solved.Based on the problems mentioned above, this article has made researches on the following aspects:1. Palm-dorsal vein recognition using Uniform Discrete Curvelet Transform (UDCT) has been proposedThe image acquisition instrument for human dorsal hand vein was developed, which was based on embedded processing system including TMS320DM642and CCD camera. The collected image was preprocessed by DSP using gray morphology method, which further improved the image quality. Then, the processed image was transmited to the PC by the Ethernet and was stored on the GUI made by Matlab for further processing. Using night vision camera constructed of another system for image acquisition. The result indicates that these systems can not only obtain the image of dorsal hand vein with high performance, but also provide an embedded development platform for hand vein recognition. Eventually, we made a database which contained400hand vein images.2. Palm-dorsal vein recognition using Uniform Discrete Curvelet Transform (UDCT) has been proposed With the palm-dorsal Region Of Interest (ROI) extraction and image preprocessing, UDCT of the Curvelet Transform was used on ROI. Then, the Curvelet coefficients phase variance was encoded. Finally, using the Chi-square distance of histogram of the coding for vein recognition obtained the recognition results. Experimental results show that this proposed method can identify palm-dorsal vein images with high robustness and high speed even by using poor-quality images.3. Dorsal hand vein recognition based on the feature fusion of UDCT phase feature and Triangulation has been proposedAfter palm-dorsal image preprocessing and ROI extraction, we use UDCT of the Curvelet Transform on ROI, and encode the Curvelet coefficients phase variance, and evaluate the Chi-square distance L1, of coding histogram for vein recognition. When the distance between L1, and threshold is too small, we detect the skeleton of the hand vein, and use the ending point and the crossing point of the extracted vein skeleton as the feature point, then measure the triangle side by the Triangulation algorithm, get L2. At last, we calculate the matching distance L by the weighted average method. This approach improves recognition ratio and avoids the cost time noise problem increase sharply.4. Palm hand vein recognition based on the feature fusion of UDCT and Log-Gabor phase feature has been proposedThe approach uses UDCT and Log-Gabor filter phase encoding to represent the texture feature of the vein image, and fuses the two coding matrices and make histogram statistics for it, then uses Chi-square distance to evaluate the matching degree. The experimental results based on the database provided by Polytechnic University (PolyU) are given and analyzed. |
PDF Full Text Request