Research of Dual-Mode Recognition Algorithm Based on Fingerprint and Finger Vein
|School||Harbin Engineering University|
|Course||Pattern Recognition and Intelligent Systems|
|Keywords||Fingerprint recognition Finger vein recognition Multi - modal identification system Image segmentation Image Enhancement Feature extraction|
Biometric identification technology is a technology that rely on biometric authentication has not lost, forgotten at the same time it is difficult to forgery and counterfeiting, etc., is gradually replacing the traditional means of authentication become the first choice for people's daily life authentication method. At present, the biometric system, mainly in the study of single-mode identification system, also achieved a lot. However, the practical application of the system also found that many of the problems, and these problems can not be effectively addressed only on the basis of the single-mode system. Multi-modal system combines multiple features to form a new identification system to improve system accuracy, noise immunity, anti-counterfeit attack and reduce the degree of attenuation of the large library, etc., which can effectively make up a single modal identification deficiencies of the system, making the new system authentication, recognition is faster and more effective. We build and to achieve a bimodal system based on fingerprints and finger vein, the system includes the following three parts: a fingerprint identification module, finger vein recognition module, and a multi-modal system integration module. Fingerprint recognition module, and the fingerprint image is normalized gray, segmentation, enhance, binarization and refined a series of pre-processing operations, making the images into high-quality binary thinned fingerprint image , to facilitate the feature extraction operation. Then, the 8-field method to refine the image feature extraction, the link of the image matching, a new fingerprint image feature point matching algorithm is proposed to complete the match, identification, and the experimental results of the corresponding analysis. In finger vein recognition module, the analysis of the finger vein image orientation field, the direction filtering algorithm based finger vein extraction algorithm, this algorithm in terms of smoothness and connectivity, or remove noise and false feature are better than the traditional method of Niblack, laid the foundation for subsequent feature extraction and recognition. Finger vein recognition process, based on the the MHD distance of the matching algorithm to complete the identification, found that the algorithm of translation, rotation serious image recognition effect is not very good, and finger vein based on relative distance matching algorithm experiments were carried out, effectively overcome the impact on the recognition result of the translation, rotation, etc., and allows the system to identify the effect of the improvement. In a multi-modal system integration module, first introduced the key technologies to build multi-modal system, and then proposed a new fingerprint image quality assessment and finger vein quality assessment method, based on based on fingerprint quality evaluation and finger venous quality evaluation decision level fusion, and fusion results were analyzed to verify the overall performance of the multi-modal system stronger than a single fingerprint and finger vein recognition system; also introduced based on the decisions of the two decision-making level. fusion algorithm, the algorithm is also better than a single identification system, the algorithm introduced in the third classifier design feature fusion concept built based on feature level fusion of multi-modal system laid the foundation for the future in-depth study.