多特征模糊推理的虹膜识别算法 an iris recognition algorithm based on multi-feature fuzzy inference.pdfVIP

第 36 卷第 5 期 光电工程 Vol.36, No.5 2009 年 5 月 Opto-Electronic Engineering May, 2009 文章编号：1003-501X(2009)05-0104-07 多特征模糊推理的虹膜识别算法 苑玮琦，张 雷，柯 丽 ( 沈阳工业大学 视觉检测技术研究所，沈阳 110178 ) 摘要：为了更加完备地描述虹膜的纹理特征，利用虹膜图像灰度信息提取局部特征点、局部纹理方向和局部纹理 的亮暗变化三种特征共同构成纹理的特征空间，克服了之前的多数虹膜识别算法提取单一特征易受干扰影响的局 限性。然后，通过设计的模糊推理规则进行模式的分类，这种分段线性分类器的设计提高了算法线性分类的能力。 分别在两个图库上进行了实验，识别率分别达到 99.41%和 99.67% ，实验数据表明：结合多种特征能较好的反映 虹膜的纹理变化特征，提高了虹膜识别的正确率，使算法具有非常优越的识别性能。 关键词：生物特征识别；虹膜识别；特征提取；模式分类；模糊推理 中图分类号：TP391.41 文献标志码：A An Iris Recognition Algorithm Based on Multi-feature Fuzzy Inference YUAN Wei-qi ，ZHANG Lei ，KE Li ( Computer Vision Group, Shenyang University of Technology, Shenyang 110178, China ) Abstract: To get more representative iris features, three features, local feature point, local texture direction and changes between brightness and darkness of local texture are extracted through the gray-scale of iris images, which can depict the feature space of texture more fully and overcome the limitation of most previous algorithms which extract only susceptible single feature. Then patterns are categorized by the designed fuzzy inference regulation. The design of this piecewise linear classifier enhances the ability of linear classification of the algorithm. Experiments are implemented in two databases, respectively. The correct recognition rate is 99.41% and 99.67%, which demonstrate that many kinds of features can represent the variation details in the iris patterns properly. Therefore, the correctness is improved and the algorithm wins predominant recognition performance. Key words: biometrics; iris reco