A Study And Realization Of Partial Fingerprint Based Mosaicking Technology For Mobile Devices

Biometric recognition refers to the use of distinctive anatomical and behavioral characteristics for identity authentication.Compared with other biometric systems,the fingerprint recognition system is more convenient and has better performance in terms of cost and performance,which makes it widely used in various fields.Among the many applications of fingerprint recognition,the most concern is its rapid development on mobile devices.In order to facilitate the embedding of the sensor in the device while reducing costs,the sensor is usually made very small.However,the fingerprint obtained by using such a small-area sensor is only a part of the complete fingerprint.It contains relatively little fingerprint information,and the overlapping area between multiple partial fingerprints is also small,resulting in a reduction in the recognition performance.Therefore,the performance of the fingerprint registration has a crucial influence on the reliability of the fingerprint identification system using a small area sensor.In order to solve this problem,the concept of fingerprint mosaicking is introduced,and the partial fingerprint images acquired during the registration phase are spliced to obtain a comprehensive template,thereby improving subsequent matching performance.This paper has conducted some in-depth researches on partial fingerprint mosaicking algorithms.The main research contents are as follows:The partial fingerprint mosaicking algorithm proposed in this paper mainly includes two parts:partial fingerprint alignment and partial fingerprint fusion.At the stage of partial fingerprint alignment,a course-to-fine alignment process was proposed.Firstly,a partial fingerprint matching algorithm is used to calculate the matching parameters of the partial fingerprints(generally also called patches)to obtain a rough alignment result.Then a reference-correlation relational map containing the global information of the patches is constructed according to the matching parameters.Based on this,a new construction method of the error energy function is further proposed.This energy function is optimized to obtain the fine alignment of the patches.At present,the existing fingerprint feature fusion algorithms are all based on minutia feature fusion,but the limited minutiae in the patches greatly limit the matching performance.In order to improve the matching performance,patch-based fingerprint matching needs to introduce additional non-minutia features,which introduces the non-minutia feature fusion problem of patches.To solve this problem,this paper proposes a novel feature fusion method,it adds ridge point fusion based on the fusion of minutiae.Firstly,according to the reference-correlation relationship map of the alignment stage,a reasonable fusion order is proposed,then the minutiae are merged,and the relationship between the ridge points to be fused is determined by establishing the connected ridges and related ridges structures for the minutiae.Thus a new ridge point fusion method is obtained.Through the combination of minutia and ridge points in this paper,a comprehensive template containing rich fingerprint features is obtained.The FVC2002 database is used to verify the algorithm in this paper,which simulates the fingerprint registration process of mobile devices.The experimental results show that the proposed algorithm has a very good performance.