Partial Fingerprint Recognition Based On Multi Method Fusion

With the digital information age approaching,authentication becomes even essential.and biometric identification technology also becomes more and more popluar.Biometrics is a technology that uses the physiological characteristics of the human body to carry out automatic identification of individual identity.Compared with the traditional identity authentication technology,such as input password,it has high accuracy,high robustness and is not easy to forge.Among them,the fingerprint recognition technology because of his high accuracy,easy to collect and so on,has gradually become the individual identity verification widely used technology.In the existing fingerprint recognition methods,the minutia of the fingerprint is the most widely used and most reliable feature.When the fingerprint image is of good quality and can extract enough detail information,it can obtain good recognition result by using the minutiae information,and the precision is very high and very mature.However,in most cases,fingerprint images are incomplete,which will affect the performance of fingerprint recognition.This is the most important meaning of the problem of incomplete fingerprint recognition.When the size of fingerprint image is smaller and smaller,the detail point of fingerprint image will be greatly reduced,which will greatly affect the reliability of authentication.With the development of science and technology,the resolution of the fingerprint picker has also been greatly leap,which allows us to collect the fingerprints on the fingerprint similar to the third layer of fingerprint features such information.In this paper,we propose a matching algorithm based on the third layer fingerprint feature-sweat hole matching algorithm and dense sift feature extraction algorithm.In the matching algorithm based on the sweat hole,the method of extracting the sweat hole is introduced emphatically.Firstly,the fingerprint image is divided into blocks according to the direction and thickness of the lines,and the model is used to extract the sweat pores.Then,the support vector machine is introduced.At the same time,the candidate regions of the fingerprint image are classified and the perspiration holes are extracted.Followed by the two methods of fusion,in the perspiration extraction to ensure high accuracy and recall rate.Finally,based on the extraction of sweat pores,the matched fingerprints were used.In the matching algorithm based on the dense sift feature extraction,after many attempts,it is found that the dense extraction of the sift feature according to the preset interval and the matching can also achieve very good results.Due to the problem of fingerprint library copyright,this paper analyzes,compares and analyzes several methods in PolyU HRF Database,which is provided by Hong Kong Polytechnic University with 1200 dpi HD fingerprint library.Experimental results show that the proposed method of sweat extraction in fingerprint matching test has a more prominent performance,while after the integration of the fingerprint alignment algorithm also maintained a relatively high level.