Close Non-matching Study On Delta Region Of Whorl

One important task of fingerprint identification is to identify a criminal suspect by comparing the fingerprint extracted from the crime scene with another one of known source stored in the automatic fingerprint identification system,then experts review these two similar fingerprints to decide whether they came from the same source.As it has the characteristics of “different from others and unchanged for life”,fingerprints has been long time regarded as the first important evidence and became an significant way to identify suspects.In practical work,by the influence of the contact mode and other objective conditions,fingerprints usually were partial so it was more difficult to identify and leads to false identifications.On the other hand,the movement by individual touching objects determines that the fingerprint delta region is more easily to be found on the object.Therefore,most of the fingerprints extracted by police investigators at the crime scene include the delta region.However,the complex flow direction of ridges and the highly frequent appearance of features in delta region,which leads to a high degree of similarity between close non-matching fingerprints,which are easily cause interference for experts.This paper focused on the influence of the quantity and quality of features at whorl's delta region on the rank of homologous fingerprint in AFIS system in million grade database;found the appearance and distribution rule of close non-matching phenomenon in delta region;analyzed and summarized the probability of occurrence of highly close non-matching fingerprints and the influence of them on AFIS ranking and experts' identification decisions under the change of quality conditions;improved the rank of homologous fingerprint in the candidate list by improving the traditional algorithm.The specific research contents included: 1.Researched the delta region of whorl fingerprint.Under the condition that the numbers and methods of marking feature points were different,how rank of homologous and close nonmatching fingerprints in the list of AFIS,and summarized rules of appearance for matching and close non-matching fingerprints.2.Researched the delta region of whorl fingermark.Under the condition that the numbers and methods of marking feature points were different,how was the probability of occurrence of homologous fingerprints,improved the understanding of features in delta region by identifiers,and gave suggestions on methods of labeling features,so to improve the rank of homologous fingerprints.3.Focused on the influence of highly close non-matching fingerprints on the ranking in AFIS candidate list under the condition of different fingermark quality,lightened the attention of researchers to improved algorithm.4.Focused on the influence of highly close non-matching fingerprints on identification results,analyzed the subjective and objective causes of identification errors,and warned the appearance of close nonmatching fingerprints and their affects on fingerprint analysts.5.Fused similar triangle matching algorithm with SIFT feature which brought STSF algorithm to improve the traditional algorithm,so that the ability of recognition algorithm to conjunctively utilize the features on incomplete fingermarks and the rank of homologous fingerprint was improved.Results of this research indicated that: 1.the number of features could affect the occurrence rate and the rank of both homologous and close non-matching fingerprints.The overall trend was with the increase of the number of features,the rank rate of homologous fingerprints increased,while close non-matching fingerprints decreased.2.If the ridges on fingermark was clear and the number of features was relatively enough,automatic feature labeling method should be adopted to benefit the homologous fingerprint,but when ridges on fingermark was not clear and the number of features was few,then manual feature labeling method and changing the features combination should be adopted.3.The occurrence rate of highly close non-matching fingerprints in the candidate list was 1.5‰,and the decrease of fingerprint quality would affect the occurrence rate of both homologous and close non-matching fingerprints.4.Fingerprint analysts should follow the procedure of identification strictly,should not be overconfident,and fingerprint identification experiments should be carry out regularly to improve the recognition of close non-matching fingerprints.5.The proposed STSF algorithm was better than traditional algorithms on identifying incomplete fingermarks,improving the rank of homologous fingerprints in candidate list,decreasing the rank of highly close nonmatching fingerprints in candidate list,and decreasing the interference caused by close nonmatching.Results of this research could provide basic data for further study on the delta region of whorl and reference for querying homologous fingerprints.It would be benefit for analysts to improving the ability of identifying close non-matching fingerprints,and decreasing the risk of false identifications.At the same time,through the improvement on traditional algorithm of partial fingermarks,the ability of AFIS to identify partial fingermarks and the rank of the homologous fingerprint in candidate lists were both improved.