The Impact of Database Size on Fingerprint Database Searches
In recent years,the rapid growth in the capacity of AFIS(Automatic Fingerprint Identification System)databases has led to an increasing difficulty in fingerprint identification,particularly in the interference of Close Non-Matches(CNMs)with homologous fingerprints in the search results.Before using AI fingerprint recognition algorithms,CNMs with higher scores and higher ranking s may appear in the candidate list.In order to explore the influence of different AFIS database sizes on the occurrence of homologous fingerprints and CNMs under the condition of traditional comparison algorithm,this experiment established 6-million-people,10-million-people,and 100-million-people level databases by setting the fingerprint card imprinting time during querying,then initiated querying and discussed the search results of each part of the delta area of the loop(root part,center part,and periphery part),and the annotations of each part were the 10 minutiae closest to the apex of the bottom-type line.The results show that when the capacity of the fingerprint database grows,the occurrence rate of homologous fingerprints decreases,and their ranking decreases at the same time,and the larger the size of the growth of the fingerprint database capacity,the more obvious the degree of decrease.When the capacity of the fingerprint database grows,the number of occurrences of CNMs increases,and the number of corresponding points of CNMs also increases,and the larger the scale of the growth of the fingerprint database,the more obvious the degree of increase.In this experiment,three high-level CNMs with 10 corresponding points were found in the 10-million-people and 100-million-people level databases.When CNMs are ranked before homologous fingerprints,it may cause interference to fingerprint examiners.In addition,it was also found that the number of occurrences of CNMs in the three parts of the delta area of the loop in different databases showed that the root part>the center part>the periphery part,which was related to the density of the minutiae in the three parts.The higher the density of the minutiae,the smaller the distance between the minutiae,and the smaller the area of distribution of the unit number of the minutiae,the higher the probability of the repetition of the same distribution pattern,and the lower the specificity of the minutiae configurations,the easier it is to produce feature similarity.This study aims to improve the risk awareness of fingerprint examiners under big data conditions.In addition to being cautious,the industry may need to do a lot of work from upgrading fingerprint matching algorithms and establishing new fingerprint identification paradigms.
万方数据
维普
