警用辨识系统现场指纹分类与点模式匹配算法研究  

A Scene Fingerprint Classification and Point Pattern Matching Algorithm Research in Police Identification

作　　者：刘元宁[1] 苑森淼[1] 朱晓冬[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130025

出　　处：《计算机研究与发展》2003年第6期813-819,共7页Journal of Computer Research and Development

基　　金：长春市科委重点资助项目 ( 990 11)

摘　　要：尽管验证模式 ( 1∶1方式 )指纹识别产品可在市场看到 ,但人们对警用辨识模式 ( 1∶N方式 )识别挑战的追求从未停止过 ,而这恰是指纹识别理论与应用领域极为重要的部分 基于K Karu等的core和delta分类模型 ,追加了严格的约束条件与相关概念定义 ,从理论上阐明了其正确性 ,同时 ,分析指纹的特征空间结构 ,提出了登记条件、匹配支持、基点对等概念 ,实现了具有旋转、平移、缩放不变性的快速匹配 5 0 0 0枚样本空间的现场指纹实验结果 。Although the verification (1∶1 mode) model' fingerprint products can be found in market, the identification (1∶N mode) model' fingerprint system in police is just what to pursue for a long time, which is of much importance to fingerprint identification theory and application field. Based on the classification model put forward by K.?Karu, adding strict restrict conditions, concepts, and definitions, this paper demonstrates its correctness. And feature space structure of fingerprint is analyzed and the concepts of registration conditions, matching support, and maximum matching pair are presented, completing a fast and precise matching with invariability of rotation, translation, and scale. The scene fingerprint experiments with 5000 fingerprints have verified the validity and correctness of the above algorithm.

关 键 词：指纹分类 模式匹配 异常点 登记条件 基点对 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]