The Study On The Partial Fingerprint Mosaicking Algorithm Based On Dual-descriptor Registration And Ridge-nearest Correction

Fingerprint identification technology,as a key technology for identity authentication,is widely utilized in many applications.The miniaturization of fingerprint collectors becomes a trend due to the increasing maturity of integrated technologies and the popularity of electronic devices.Thus,fingerprints collected are much smaller than conventional ones,resulting in many difficulties in the process of fingerprint matching and recognition.The mature fingerprint matching or recognition algorithm has a sharp decline in performance on partial fingerprints.Therefore,the research of partial fingerprint identification has become one of the hot spots in this field.Partial fingerprint mosaicking is carried out during the registration phase.Two or more fingerprint images are transformed into a composite fingerprint image through a series of spatial transformations.Then the composite fingerprint image is used as a template for recognition to improve the efficiency of partial fingerprint recognition.Research shows that fingerprint mosaicking technology is one of the effective ways to improve the accuracy of small area fingerprint recognition.Traditional fingerprint mosaicking algorithm uses rigid transformation to mosaic two fingerprint images.Alternatively,it uses the fingerprint minutiae pair as the land marks for adjusting the deformation of the fingerprint image.The first kind of fingerprint mosaicking algorithm selects the minutiae point as the mosaicking datum point,and then it uses rigid transformation to mosaic the fingerprint image without considering the deformation.The second one considers the existence of deformation.However,the deformation parameters rely excessively on the minutiae.Partial fingerprint image has few details,and the false minutiae have a great influence on the algorithm.This article combines the relevant knowledge of image processing and fingerprint mosaicking to conduct in-depth research in the area of partial fingerprint mosaicking.The main contributions of this article are as follows:(1)For the problem of lack of sufficient minutiae in partial fingerprint images,a fingerprint mosaicking algorithm based on dual-descriptor registration is proposed.In the premosaicking phase,we combine the matching scores of two typical fingerprint minutiae descriptor.The proposed algorithm establishes more relationship between the single minutia with its surrounding information,which can improve the reliability of the results of partial fingerprint mosaicking.(2)For the pre-mosaicking parameter selection problem,we use the steepest descent method to optimize the mosaicking parameters based on the ridge mosaicking error of the distance image,which not only improves the mosaicking quality,but also improves the efficiency.(3)On the basis of rigid fingerprint mosaicking,the k-neighborhood block method is used to obtain the paired ridges in the overlap region of the fingerprint image.The TPS model is used to correct the elastic deformation of the fingerprint image from the ridge line,which eliminates the noise impact caused by rigid mosaicking algorithm.(4)To address the color difference problem resulting from gray scale inconsistency in mosaicked image,we propose a weighted smoothing algorithm for image fusion which can eliminate mosaicking seams.This method considerably improves the quality of mosaicking and visualization,and it is capable of improving the recognition rate of fingerprints.In this paper,experiments are performed on the self-built XDFinger partial fingerprint database.We analyze the performance of partial fingerprint mosaicking algorithm based on dual-descriptor registration and ridge neighborhood correction.Firstly,we compare the mosaicking results of the rigid mosaicking and the mosaicking image after correction.The effectiveness of the mosaicking algorithm is illustrated from the perspective of visual perception.Secondly,the minimum distance,maximum distance,distance average and distance variance indicate the accuracy of the mosaicking algorithm.The data verify the accuracy of the splicing algorithm.Finally,in the recognition performance,the EER of the rigid mosaicking algorithm of this paper reaches 0.95%,which is significantly lower than the 1.64% and 1.33% of the traditional rigid mosaicking algorithms,and the correction algorithm EER reduces the recognition algorithm EER from 0.89% to 0.68%.Experiment results show that the fingerprint mosaicking method in this paper improves the mosaicking quality and recognition performance of partial fingerprint images.