Research On Palm Vein Recognition Methods

As the society advances constantly,network technology has developed rapidly and people's security awareness has been increasingly intensified.In the context,the conventional authentication method has been far from fulfilling people's requirements in the dimensions of convenience,reliability and security.Given the circumstances,biometric recognition technology,which makes full use of human's intrinsic physiological characteristics and extrinsic behavioral characteristics,has emerged as a powerful and effective alternative.Among them,palm vein stands out of myriads of intrinsic biometric features,which is a promising method of recognition.In terms of palm vein recognition technology,a complex and meaningful topic of research has been how to implement a highly accurate,highly reliable and highly efficient method of palm vein recognition.The main work of this thesis is listed as follows:(i.)The biometric recognition technology is elaborated in the thesis,and the common biometrics technologies such as fingerprint,palmprint and iris are presented.Next,the author comparatively studies these recognition techniques and palm vein recognition technology to highlight the advantages of the palm vein recognition technology.Then,the recognition process and recognition methods of palm veins are summarized.The operation mode of the palm vein,the method of noise reduction in the ROI region of the palm vein and the performance evaluation indexes are explored.The common database of palm vein is also presented in the thesis.(ii.)Based on mutual texture local derivative pattern(MT?LDP),an algorithm for feature extraction of palm vein is proposed in the thesis.Firstly,the study adopts the maximal principal curvature algorithm and k-means method for texture extracting,which effectively suppresses the noise and improves the accuracy and robustness of texture extraction.Secondly,according to the texture extraction results,the mutual texture(MT)image is synthesized.The mutual texture image consists only of the palm vein texture and the grayscale image of its neighborhood,which contains most of the information useful for recognition while excluding the background information,thus reducing the interference.Additionally,the maximal matching pixel method is employed to find the best matching region of the two textures for comparison,which further enhances the performance of the LDP algorithm.Finally,the LDP features are extracted from the MT images on the basis of LDP algorithm,and the matching degree of the two MT-LDP images are calculated by the improved Chi-square distance.(iii.)The palm vein feature extraction algorithm is raised based on Euclidean distance and Cosine distance SIFT.By investigating the scale space and SIFT algorithm,it is found that SIFT algorithm may lead to some error matching points in the course of palm vein recognition.In view of the large number of error matching points in SIFT algorithm in feature points matching,an improved solution is pointed out in the thesis.The error matching points are firstly eliminated with the RANSAC method,and then further purified by Euclidean distance and Cosine distance.After the improvement of SIFT algorithm,better recognition results have been obtained in many databases of palm vein.