基于特征融合和优化极限学习机算法的虹膜识别系统  

AN IRIS RECOGNITION SYSTEM BASED ON FEATURE FUSION AND OPTIMIZED EXTREME LEARNING MACHINE ALGORITHM

作　　者：路春辉[1] 

机构地区：[1]广东工程职业技术学院信息工程学院,广东广州510520

出　　处：《计算机应用与软件》2016年第7期326-333,共8页Computer Applications and Software

基　　金：广东省科技厅项目(粤科规财字[2015]72号)

摘　　要：针对虹膜图像采集过程中存在光照、干扰等因素,为进一步提高虹膜图像识别的准确率,提出一种基于组合特征提取的优化极限学习机(ELM)模型来提高虹膜图像识别的精度。模型考虑了特征提取和分类器优化两者均起着重要作用,利用灰度共生矩阵(GLCM)和多通道2D Gabor滤波器特征提取后进行特征融合,得到更丰富的特征信息,并设计改进了蜂群算法(IABC)优化ELM模型作为分类器。同时设计的线性加权多目标函数综合考虑分类精度和网络结构,从而有效提高了虹膜识别的准确率。实验表明提出的模型通过结合两种特征提取方法,能提取出更丰富的可区分特征,并且结合优化分类器得到了很高的分类准确率,是一种有效的虹膜识别模型。The acquisition of iris image is easily affected by the uneven light and noise interference,thus,an optimized extreme learning machine( ELM) based on feature extraction is proposed to improve the accuracy of iris recognition. In the proposed model,considering both feature extraction and classifier selection play an important role,the gray level co-occurrence matrix( GLCM) and multi-channel 2D Gabor filters are used for separate feature extraction,and then two groups of feature sets are combined for achieving more information characteristics.Moreover,an improved artificial bee colony( IABC) is designed to conduct parameter optimization for ELM. Besides,a linear-weighted multi-objective function is designed to take into account the average accuracy rate and the number of ELM network hidden-layer nodes,which helps to improve the accuracy of model. Experimental results show that the proposed algorithm not only extracts more distinguished information by the combination of two feature extraction methods,but also obtains the highest accuracy,which proves the validity of the proposed model.

关 键 词：特征提取 虹膜识别 灰度共生矩阵 多通道2DGabor滤波器 极限学习机 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]