摘要
近年来,随着虹膜识别技术在公安领域的大规模应用,虹膜鉴定方法研究成为了法庭科学领域的研究热点。虹膜图像中可用于鉴定的特征主要有块状特征、线状特征和环状特征。本文以虹膜块状特征为研究对象,对异源虹膜块状特征的相似性进行统计分析。首先对虹膜图像做归一化处理,并利用Tophat算法消除背景噪声,进而提取块状特征;然后基于Hu矩特征和位置坐标信息对块状特征的相似度进行描述;在952张虹膜图像的数据集上进行统计分析,统计结果表明,两枚不同的虹膜上1个块状特征相似的比率为4.38%,2个块状特征相似的比率为0.534%,3个块状特征相似的比率为0.0437%,4个块状特征相似的比率为0.00353%。本文提出的异源虹膜块状特征相似性的统计方法,可用于计算大规模虹膜数据集下异源虹膜间出现不同数量相似块状特征的概率,为解决鉴定过程中需要多少个块状特征相同才能认定同一的问题奠定基础。
The rapid development of iris recognition technology has been bringing iris into becoming an ever-emerging biometric feature for identity verifi cation.The recent application of iris recognition in public security has even delivered lots of relevant researches for forensic science.Comprehensively,the current iris identifi cation builds on solving the following two problems:1)What features are eligible of the basis for iris identifi cation;2)How many iris features are required for iris identifi cation.As iris texture is divided into three characteristic shapes:block,linear and ring,thus the iris identifi cation is decided with these shapes.For the second problem,this paper tried to conduct a research into the block-shaped iris feature to statistically analyze the similarity among the heterogeneous origins of such feature.Firstly,the iris image was collected and normalized through elimination of the background noise by tophat algorithm,hence having the block-shaped feature extracted.Successively,the block-shaped features were described of their similarity based on Hu-moment feature and position-oriented coordinate information.Statistical analysis was followed to carry out into the gathered 952 sets of iris image data.The results showed that there was a likelihood of 4.38%/0.534%/0.0437%/0.00353%for two different irises to simultaneously appear one/two/three/four similar block-shaped feature(s).Depending on the capacity of the iris database,the above four likelihoods could be of small or large values.If more parameters were to describe the feature shape from multiple angles plus more experimental data,the statistical probability would be more accurate and actual.The method proposed in this paper is of value about how many iris features should be taken to distinguish identity for identifi cation purpose.
作者
张翌阳
唐云祁
陈子龙
苗迪
ZHANG Yiyang;TANG Yunqi;CHEN Zilong;MIAO Di(Investigation School of People’s Public Security University of China,Beijing 100038,China;Institute of Forensic Science,Ministry of Public Security,Beijing 100038,China;Beijing IrisKing Co.Ltd.,Beijing 100190,China)
出处
《刑事技术》
2023年第1期53-61,共9页
Forensic Science and Technology
基金
国家自然科学基金项目(61906199)
中国人民公安大学公共安全行为科学实验室开放课题(2020SYS04)。
关键词
虹膜鉴定
块状特征
相似度
统计分析
iris identifi cation
block-shaped feature
similarity
statistical analysis
