基于多子区域联合的高适应性虹膜识别算法  

An Iris Recognition Algorithm of High Adaptability Based on Multiple Region Combination

作　　者：苑玮琦[1] 张雷[1] 柯丽[1] 

机构地区：[1]沈阳工业大学视觉检测技术研究所,辽宁沈阳110178

出　　处：《电子学报》2009年第5期981-986,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60672078;No.60472088)

摘　　要：由于在采集虹膜图像前,无法预知眼睑、睫毛等噪声对虹膜纹理的干扰程度和不受干扰的可用虹膜区域的位置和大小,这可能会使提取到的特征模板中包含了由噪声引起的不可靠和不稳定特征,使识别的错误率增加.本文提出了多子区域联合的识别方法,将相对不易受干扰的图像区域划分为4个子区域,分别计算两幅图像对应子区域的相似度,动态选择最相似的子区域,将其特征作为判定依据进行分类.克服了之前算法只选择一个固定位置的区域用于特征提取的局限性.采用CASIA虹膜图库进行测试,结果表明:本方法能提高识别准确率、增强算法对采集图像质量要求的适应性,改善了虹膜识别系统的性能.As eyelid and eyelashes are likely to disturb iris texture, the extent cannot be predicted before acquiring iris image. Some fluky and unfixed features are caused by the yawp in feature templates, which make iris classification false rate increase. In order to resolve this problem, a iris recognition algorithm based on multiple region combination is proposed in this paper, iris image that is not liable to be disttnbed relatively is separated into four subarea,then the four similarities of corresponding subarea are calculated in two image,features of the closest subarea are considered as the judgment that is used to carry out classification.It overcomes previous limitation brought by selecting only a fixed position to extract features. Results show that proposed means achieves quite high accuracy, it is efficient for boosting up adaptability to image quality and improving iris recognition performance on CASIA iris database.

关 键 词：生物特征识别 虹膜识别 特征提取 匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]