分块统计特征和最优分辨力选择特征相融合的掌纹识别  

Palmprint Recognition Based on Block Statistic and Optimal Resolution Criterion Selection Features

作　　者：林红[1] 孙雅娟[1] 

机构地区：[1]华北电力大学网络与信息中心,北京102206

出　　处：《科学技术与工程》2014年第3期212-215,221,共5页Science Technology and Engineering

摘　　要：为提高掌纹识别的性能,提出一种分块统计特征和最优分辨力选择特征相融合的掌纹识别方法。首先对预处理后的掌纹图像进行多方向、多尺度Gabor变换;然后将掌纹划分多个子块提取特征,将各子块特征进行拼接得到整个掌纹特征向量;最后以特征分辨力为准则选出最优掌纹特征子集建立两分类器,通过投票机制建立掌纹多类分类器,并采用Po1yU掌纹库进行性能测试。测试结果表明,该方法的掌纹识别性能优于对比掌纹识别方法。In order to improve the performance of palmprint recognition, a novel palmprint recognition method based on block statistic and optimal resolution criterion selection features is presented. Firstly, the palmprint image is preprocessed by gabor transform, and then the palmprint image is divided into blocks, and features of palmprint image are extracted, the features were assembled to form the palmprint image feature vectors. Finally the optimal features are selected according to feature resolution criterion to establish two classifier, and the palmprint multi- class classifiers is established by voting mechanism, and the performance is test by PolyU palmprint database. The results show that, compared with the traditional palmprint recognition methods, the proposed method has improved the palmprint recognition rate.

关 键 词：掌纹识别 GABOR滤波 分块统计 特征提取 投票机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]