基于改进K-Means的静脉特征学习与识别  

Hand Vein Feature Representation and Recognition with Improved K-Means Model

作　　者：孙伟[1] 刘晓敏[1] 王浩宇[1] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008

出　　处：《控制工程》2017年第9期1751-1755,共5页Control Engineering of China

基　　金：中央高校基本科研业务费(2013XK09)

摘　　要：针对传统的静脉识别中静脉特征的提取需要通过先验知识进行人工设计,并且特征的设计过程中需要对大量的参数进行调整,同时需要在后续的分类器设计中进行特殊的选择才能达到较好的识别效果等缺陷,提出了一种改进方法,对单层网络的特征学习结构中的K-Means方法进行针对性改进,并将其引入到静脉识别的静脉特征学习过程中,在分类器中采用SVM实现静脉分类。另外,引入SIFT特征结合改进词袋模型(SBOW)的传统特征学习和分类方法分别进行静脉识别,并将两者的识别结果进行对比,从而证明将基于单层网络特征学习方法引入静脉识别中的优越性和必要性。Performance of feature extraction and representation, sticking point of image recognition task, will directly influence the accuracy of final recognition. The traditional feature representation method of vein recognition is based on the sufficient prior knowledge of analysis on vein information characteristics, the shortcoming of which reflects in long time consumption spent on scheduling parameters and special selection about later classifiers to guarantee the final recognition rate as high as possible. The paper makes the attempt to introduce the K-means model and single-layer feature representation architecture to the vein recognition task with some targeted modifications, and adopts the SVM at the link of classifier design. Meanwhile, the paper also introduces the ＂bag of features＂ model combining the SIFT feature to realize the vein recognition experiment design, and by realization and comparison of the two methods, the paper tells that why the importation of feature representation methods to vein recognition is necessary and which feature representation is exactly the better one in the vein recognition task.

关 键 词：静脉识别 单层网络特征学习 K-MEANS SIFT特征 词袋模型 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]