摘要
提出了一种最优轴投影(OAP)特征来描述步态信息。首先用背景减除法得到序列图像中的轮廓,然后随机选一部分轮廓样本基于主成分分析(PCA)方法训练出最优轴,并基于最优轴提取每个轮廓点的OAP特征。在UCSD和SOTON数据库的实验结果显示,OAP特征包含了丰富的步态信息;和同类型方法相比,本文方法有较高的识别率,计算复杂度减小,识别速度提高。
A new gait feature called as optimalaxis projection（OAP） is proposed as gait information in this paper.First,the background subtraction is used to extract silhouettes from the regional image sequences.Then we randomly select some silhouette sequences to train the optimal axis based on the principal component analysis（PCA） method,and extract the OAP features through the optimal axis.Experiments on UCSD and SOTON database show that the OAP features contain abundance of gait information.Compared with the same type methods,the OAP method can reach a higher recognition rate.Furthermore,the method reduces the complexity of computation and consumes less recognition time than other methods.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2010年第12期1856-1859,共4页
Journal of Optoelectronics·Laser
基金
教育部新世纪优秀人才支持计划资助项目(NCET-08-36)
