Research On Fingerprint Quality Evaluation And Matching Algorithm

With the natural advantages of lifetime invariance and high discrimination,fingerprint recognition becomes the most widely utilized biometric recognition technology and is easy to implement in many areas,also it produces good social benefits.In view of degradation of fingerprint recognition performance due to fingerprint quality problems occurred in many occasions,and low performance stability using a single fingerprint recognition algorithm,based on sorting and collecting a large number of documents,data fusion based fingerprint quality evaluation and fingerprint cascade matching method are proposed in this research.The research and innovation are carried out from the following two aspects:Firstly,data fusion based fingerprint quality evaluation is proposed to improve the performance of the fingerprint matching.As a method for processing mess data,data fusion is utilized for combining multiple decision results to effectively improve the fingerprint judgment effect.The quality problems include several aspects:Fingerprint image acquisition area is small;The fingerprint image position is deviated;Fingerprint moisture is uneven;Fingerprint is not clear.The first step is to get five quality scores,and then a supervised learning algorithm is used for the integration of the five quality scores.Through the experiment,it can be proved that the accuracy of the method can be improved by fusing the information of multiple indexes,and the effectiveness of the method is proved.Secondly,a fingerprint matching method based on cascade decision is proposed to improve the recognition performance of the matching algorithm.In view of the situation that the matching strategy of the conventional fingerprint algorithm is not satisfactory,the synthesis algorithm is proposed,which is based on three stage cascade matching:center point matching,middle region matching and global matching.For the two aligned fingerprints,the above-mentioned strategy is performed successively.And in this process,if there was any evidence showing they are the same person,the calculation will be terminated.If the last process got low similarity,which means they are not the same person.Through the design of two experiments we proves that strict quality judgment can significantly delete the image with poor quality,and cascade method could also improve the performance of fingerprint matching.