一种改进的2DPCA车脸识别方法  

An improved method of car face recognition based on 2DPCA

作　　者：燕磬[1] 杨安平[1] 陈凌宇[1] 

机构地区：[1]长沙理工大学,湖南长沙410004

出　　处：《现代电子技术》2014年第8期123-125,128,共4页Modern Electronics Technique

摘　　要：传统的2DPCA算法在识别的过程中需要计算出训练样本的平均值,在训练样本过多图像分辨率过高的情况下,无疑会使得计算时间过长,为了解决这个问题,在此提出了一种将样本数据重新排列之后提取中间的某些数求简化均值的方法,以此简化均值数重建散布矩阵。实验之后表明,在训练样本较多时且训练图像分辨率较高时,识别速度有大幅提高,且取得了较高的识别率。The traditional 2DPCA algorithm needs to calculate the mean of training samples in the recognition process. How-ever,it will spend too long time if there are too many training samples and the image resolution is too high. In order to solve this problem,a new two demension principal component analysis（2DPCA）method based on simplified mean that extracts some numbers in the middle after sampling data re-arrangement is proposed. The simplified mean is used to reconstruct the scatter ma-trix. The experiment result shows that this method can increase in recognition speed and has a higher recognition rate even if there are too many training samples or the image resolution is too high.

关 键 词：二维主成分分析 特征提取 简化均值 车脸识别 

分 类 号：TN919-34[电子电信—通信与信息系统] U495[电子电信—信息与通信工程]