基于Radon变换特征提取的步态识别  

Feature Extraction Based on Radon Transform for Gait Recognition

作　　者：张前进[1] 徐素莉[1] 孙炎增[1] 

机构地区：[1]河南科技大学电子信息工程学院,洛阳471003

出　　处：《数据采集与处理》2010年第2期268-272,共5页Journal of Data Acquisition and Processing

基　　金：河南省重点科技攻关基金(082102240086)资助项目

摘　　要：提出了一种基于Radon变换特征提取的步态识别算法。该算法根据步态轮廓图下肢的宽度信息确定步态运动准周期性,对二进制准周期步态轮廓序列进行Radon变换构造特征向量模板。对特征向量进行主成分分析,并采用k-近邻法进行步态特征分类。在CASIA步态数据库上和CAS识别算法进行了详细的比较,实验结果表明,该算法在性能上有较大程度的提高,是一种有效的步态识别方法。A novel feature extraction algorithm is proposed for the gait recognition based on Radon transform. For each image sequence, the algorithm obtains the gait quasi-periodicity by analyzing the width information of the gait contour edge of lower limbs. Radon transform is applied to the binary quasi-periodicity image sequence for feature vector constructing. The principal component analysis is used for feature vector construction, and the k-nearest neighbor classifier is adopted to recognize subjects. Compared the algorithm with the typical CAS algorithm on CASIA gait database, experimental results demonstrate that the proposed algorithm achieves a higher recognition rate.

关 键 词：步态识别 RADON变换 主成分分析 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]