Studies Of The Algorithm For Vein Far-Infrared Images Of Dorsal Hand

With the rapid development of the information technology, biometric identification technology is used widely in the field of national safety, judicature, finance and electronic commerce, etc. Up to now, fingerprint identification is the most mature identification technology. However, there exist some defects, such as being affected by the condition of finger skin, contact sanitary problem, being easily copied and forged by criminals. Therefore, a novel technique of biometric identification based on human dorsal hand vein pattern is researched in this paper to reduce the defects. As a result of using body interior characteristics, dorsal hand vein pattern recognition has the virtue of high security.Dorsal hand vein pattern recognition is a new biometric identification technology. It utilizes the characteristics of hand vein pattern to identify different persons. The most emphases and critical task is to extract the skeleton of hand vein images. Skeleton algorithms including traditional segmentation-thin, immersion watershed, morphological filtering watershed, adaptive wavelet–based watershed are studied in this paper. The analysis results illustrate that segmentation-thin algorithm is too complicated and induces too many errors, immersion watershed transform causes serious over-segmentation phenomenon, morphological filtering watershed algorithm is sensitive to structuring element, adaptive wavelet–based watershed algorithm can't keep the edge information perfectly. Therefore, an improved watershed algorithm combined with adaptive curvelet transform is proposed to skeletonize the hand vein pattern according to the anisotropy of the images to resolve the problems mentioned above. The experimental results imply that this algorithm can extract the feature of skeleton accurately and has more virtue than the previous algorithms. The algorithm also eliminates the over-segmentation phenomenon of watershed algorithm effectively and provides a new approach for image segmentation.Moreover, feasible experimental methods and the matters needed attention are summarized. In the phase of image preprocessing, the images are extracted the region of interest(ROI), denoised and normalized. A new algorithm combined with Canny operator and Freeman code is proposed to locate the ROIs of hand vein images. The experimental results indicate that the algorithm can locate accurately.Research in this paper is significant for dorsal hand vein pattern recognition. Meanwhile, it provides a new approach for biometric identification based on human vein pattern.