Research On Key Algorithms For Biometric Recognition Based On Hand Dorsal Vein

With the rapid development of information technology, information exchange volume is increasing constantly, the security of personal information is facing with serious challenges. on this occasion, the personal identification technology is asked for higher requirements. In the field of public safety, traditional identification technology has been gradually revealing its drawbacks, vein recognition as a new biometric technology, which as its unique advantages is being paid more and more attention.Through refering to a large number of domestic and foreign literatures, this thesis analysis the existing algorithms and carrys in-depth analysis and research on the key issues of vein recognition, later, this thesis proposes new ideas and methods for vein recognition, the main work includes the following aspects:Firstly, according to the characteristics of the vein image, proposing a region of interest extraction algorithm based on the maximum inscribed circle. And on this basis, calculating the rotational angle of the hand, correcting the region of interest according to the rotation angle. This algorithm is based on hand contour, there is no influence of human factors, experimental results show that this method has good robustness.Secondly, analyzing the grayscale distribution of the vein image which is waiting for segmentation, proposing a seed point selection method which is base on valley point detection of grayscale distribution. This method can detect the pixels belonged to the vein, regarding these pixels as seed point for region growth, not only can segment the vein from the image completely, but also can avoid the impact of the shaded effectively.Thirdly, analyzing the vein feature extraction algorithm, combining geometric invariant moment and multi-resolution analysis, putting forward a vein feature extraction method based on curvelet transform, and vertifing this method’s effectiveness by experiments. The experiments results show that this method meets the requirements of the vein feature extraction and the feature vector can be used for vein classification and identification.Finally, according to the characteristics of the vein database, this thesis comes up with a vein recognition strategy which is used for large-scale applications in case of small sample size and multi-class applications. This recognition strategy takes advantage of clustering, classification and matching algorithm, and combines them together, giving consideration to vein recognition system’s accuracy and real-time in the case of large intravenous database capacity.At the end of this paper, summaring the research results which are obtained, and the further research work are prospectived.