The Study On Mosaicking Algorithm Of Small Area Fingerprint Image

As the most widely used biometric recognition technology,fingerprint identification technology has been used for personal identity.However,with the rapid development of portable electronic equipment and acquisition technology,the area of fingerprint acquisition is more and more small.Thus,the feature information obtained from the acquired fingerprint image is also reduced.At the same time,the overlap area between the different fingerprint images of the same finger also decreases,which seriously affects the performance of the fingerprint recognition system.Mosaicking technology of small area fingerprint image by merging fingerprint images or feature templates that belong to the same finger to increase the information contained in the fingerprint template,effectively improve the performance of the fingerprint identification system.Therefore,the mosaicking technology of small area fingerprint image is one of the new research hotspots in the fingerprint identification technology.Mosaicking of small area fingerprint image is an important step to form a small area fingerprint template database,which is divided into two research directions: feature mosaicking and image mosaicking.The traditional feature mosaicking only uses the independent information of minutiae,ignoring the relevant information between the minutiae and the surrounding region,which may lead to a large error in the feature template.Two classic mosaicking algorithms of small fingerprint image which based on the improved iterative closest point,and the selection of reference point for initial transform in them is using multiple points to calculate the corresponding minutiae pairs,then the reference point that results in the maximum number of corresponding pairs is chosen or selecting the core point in each fingerprint image.The former has a large amount of computation,and the latter is not suitable for fingerprint images without core point.In addition,these algorithms only use the minutiae information to determine whether the fingerprint image is mosaicking accurately,thus the reliability is low.In order to improve the shortcomings of the above,the main work and innovation of this paper are as follows:1.A mosaicking algorithm of small area fingerprint feature based on orientation descriptor and rigid transform is proposed.In this algorithm,the orientation descriptor of minutiae is added in the fingerprint feature template.Then several corresponding pairs with highsimilarity of orientation descriptor are selected and calculated the transformation parameters and matching minutiae pairs corresponding to each corresponding pair.Finally,we select the transformation parameters that results in the minimum distance between matching minutiae pairs as the transformation parameters in fingerprint feature mosaicking.The proposed algorithm improved the accuracy of matching minutiae pairs and the feature mosaicking error caused by the error matching between similar minutiae points is effectively reduced.2.A mosaicking algorithm of small area fingerprint image based on distance image and minutiae is proposed.This algorithm is based on minutiae matching algorithm to get the matching minutiae pairs of two fingerprint images.And the matching minutiae pair with highest degree of similarity is selected as reference point.Then by introducing the ridge information of the fingerprint image from the distance image,the ridge matching error is used as a criterion for judging whether the fingerprint image mosaicking is accurate.The algorithm reduces the computational complexity and further ensures the accuracy of the fingerprint mosaic image.In order to verify the validity of the two algorithms proposed in this paper,we experiment with XDFinger,a small area fingerprint database built by Xidian University,which has 200 fingers and 10 images per finger.The equal error rate of the recognition system is reduced from 0.45% to 0.39% after adding the fingerprint mosaicking feature templates obtained by the proposed feature mosaicking algorithm in this paper.Using the Veri Finger fingerprint identification system,the equal error rate of XDFinger and after adding the fingerprint mosaicking image obtained by using Jain algorithm are 0.4% and 0.34%,respectively.Adding the fingerprint mosaicking image obtained by proposed image mosaicking algorithm in this paper,the equal error rate is 0.22%,significantly increasing the recognition rate,improving the user experience of smart devices with fingerprint recognition system.