NFIQ 2.0用于指纹质量控制的可行性及方法

Feasibility and Method of NFIQ 2.0 Adopting into Fingerprint Quality Control

目的测试NFIQ 2.0标准计算指纹质量评分模块在活体采集和油墨捺印指纹质量控制上的可靠性,研究利用NFIQ 2.0指纹分值综合判断指纹质量的可行性和在实际工作中的使用方法。方法一是利用NFIQ 2.0模块接口程序结合VS2015 C#编程语言制作指纹质量评分软件。二是记录20张指纹图片放置在采集框上、下、左、右、中五个位置时NFIQ 2.0的分值,计算平均偏差、最大偏差和相对偏差。三是对在指纹数据库中随机选出的1334枚活体采集和油墨捺印指纹(8位灰度、500 dpi BMP格式指纹图片)进行评分,将得出的分值与通过人工检查该1334枚指纹在指纹系统中自动提取的特征所计算的特征准确率,形成线条图并制作拟合线,通过线条图和拟合线研究NFIQ 2.0分值与特征准确率之间存在的关系。结果从平均偏差、最大偏差和相对偏差分析,指纹图片放置位置对NFIQ 2.0分值的影响很小。利用线条图和拟合线进行统计时,NFIQ 2.0分值与特征准确率之间呈现出一定的对应关系,NFIQ 2.0分值高的指纹,系统自动提取的特征准确性相对较高,反之相对较低。并且,分值在41~100之间时,特征准确率稳定在80%以上。结论NFIQ 2.0标准计算指纹质量评分模块可以用于大批量活体采集和油墨捺印指纹的辅助质量检测,在实际工作中可将NFIQ 2.0模块嵌入活体采集软件和指纹系统档卡扫描中,设定合理的阈值,对提高指纹采集的质量检测效率和提升指纹采集质量将有较大帮助。

Objective To test the reliability of NFIQ 2.0(a software module to score fingerprint quality)in quality control for fingerprints collected under living and/or stamping ink,and explore the feasibility for NFIQ 2.0-scored fingerprint to be comprehensively judged of its quality and the relevant applicative method in practice.Methods The NFIQ 2.0 module interface program was coalesced with VS2015 C#programming language to have one fingerprint-quality scoring software contrived.Twenty fingerprint images were placed onto the acquisition frame at five positions(top,bottom,left,right and middle),having their respective corresponding NFIQ 2.0 scores calculated together with their average/maximum/relative deviation.1334 fingerprint(previously stored after collected under living and/or stamping ink)images(8-bit scaling grayness,500 DPI BMP format)were randomly selected from fingerprint database,determined of their NFIQ 2.0 scores that were afterwards compared of the feature-extracting accuracy(from the accurately-extracted features to derive the NFIQ 2.0 score)against the veracious extraction through manual verification.Therefore,the line graphs were resulted and thereby the fitting lines come forth.The correlation was determined between the NFIQ 2.0 scores and feature-extracting accuracy based on the line graphs and fitting lines.Results From the analysis of the average/maximum/relative deviation,all of the five positions(top,bottom,left,right,and middle of the frame)were shown of little effect on the NFIQ 2.0 scores.Through statistics from the line graphs and fitting lines,a certain correlation was present between the NFIQ 2.0 score and the feature-extracting accuracy.High NFIQ 2.0-scored fingerprints were accorded to relatively high accuracy of features automatically extracted with the system,and vice versa,relatively low accuracy.Moreover,for NFIQ 2.0 scores between 41 and 100,the featureextracting accuracy was kept stable over 80%.Conclusions NFIQ 2.0 scoring fingerprint-quality module is able to assist in quality control for mass living collection and ink-stamping of fingerprint.Actually,the NFIQ 2.0 module can be embedded into the software for living collection and/or used with file cards scanning into fingerprint system,excelling in a reasonable threshold set to improve the detection efficiency and the quality of fingerprint collection.