刑事技术2023，Vol.48Issue(6)：623-627,5.DOI:10.16467/j.1008-3650.2023.6015
多生物特征识别匹配层融合分析在刑事技术中的应用
Application of Multi Biometric Matching Layer Fusion Analysis in Criminal Technology
冯建刚 1杨晓琳 1景冬喜 1高泽华 1贾东涛 1伊海1
作者信息
摘要
Abstract
In actual criminal casework,we often encounter some on-site physical evidence materials such as fingerprints,DNA,and portraits which cannot be useful enough to find the suspect with poor objective conditions,and often putting the investigation of cases in a difficult position.This paper explores the fusion analysis of multiple biological features for criminal inspection work,aiming to provide new ideas for deeply exploring the value of physical evidence and material.Firstly,we create experimental samples,including 1 500 on-site fingerprint samples with three patterns and 4 to 8 features,as well as 312 facial pose photo samples with 13 angles,three perspectives,and different occlusions.We send them to the fingerprint and portrait system for comparison,and calculate the recall rates(Rf and Rp)of the target sample ID under different conditions,as well as the maximum ranking Mf of the target sample ID in the fingerprint candidate and the minimum similarity Sp in the portrait candidate set.Then,the average recall rate of the fusion analysis is calculated using the theory of multi biometric feature recognition matching layer fusion analysis combined with probability formulas.Finally,we write a fusion analysis software to calculate the accuracy P of the fusion analysis results through 12 sets of serial and parallel cases.After excluding experimental samples with lower actual combat effectiveness,the average recall rate of fingerprints Rf is 71.5％,the maximum ranking Mf is 9 984,the portraits average recall rate Rp is 70.5％,the similarity Sp is 70.3％,and the average recall rate R of multi biometric recognition matching layer fusion analysis is 92％,the precision rate is 34.6％.Multi biometric feature recognition matching layer fusion analysis can quickly identify valuable investigative targets from candidate sets of criminal casework material evidence comparison with poor objective conditions,and can explore the value of them effectively and deeply.Meanwhile,these methods strengthen the correlation application of criminal technologies such as fingerprints,DNA and portraits.Its recall rate can meet the needs of practical operations with a good precision rate and barely invalid objects that need to be identified.Combining knowledge graph technology to analyze serial and parallel cases can more effectively expand the application space of fusion analysis as well as avoid the problem of a single case being unable to fuse analysis due to only one biological feature.Our study has achieved good results in practical work.关键词
多生物特征识别/匹配层融合/召回率/查准率Key words
multi biometric recognition/matching layer fusion/recall rate/precision rate分类
社会科学引用本文复制引用
冯建刚,杨晓琳,景冬喜,高泽华,贾东涛,伊海..多生物特征识别匹配层融合分析在刑事技术中的应用[J].刑事技术,2023,48(6):623-627,5.
