共查询到19条相似文献,搜索用时 64 毫秒
1.
类似于通过求解图像协方差矩阵特征值所得到的特征脸方法,提出在双三次插值得到图像降维基础上,再进行线性鉴别分析(LDA)的插值脸方法,从而得到鉴别矢量集,实现人脸图像识别,试验结果显示,插值脸方法比普遍采用的特征脸方法效果更好.所提出的思想方法对于人脸识别的理论研究和工程应用具有较大价值. 相似文献
2.
基于2DLDA方法,提出了一种基于图像分块的二维线性鉴别分析(M2DLDA)的人脸识别方法。该方法首先对原始人脸图像进行必要的预处理后进行分块,再对分块后的子图像分别采用2DLDA方法进行特征提取,最后用最小距离分类器进行识别。该方法的优点:分块后能有效的抽取人脸图像的局部特征有利于分类;降低了2DLDA方法提取的特征矩阵的维数;特征提取是基于图像矩阵的,抽取方便快速。在ORL人脸数据库上的实验结果表明:该方法在识别性能上优于2DLDA方法。 相似文献
3.
基于PCA+LDA的热红外成像人脸识别 总被引:3,自引:0,他引:3
研究热红外成像人脸识别技术,提出一种基于主成分分析(PCA)和线性鉴别分析(LDA)的热红外成像人脸识别方法.针对热红外人脸图像的特点,首先对图像进行预处理得到一组标准热红外人脸图像,利用PCA算法对图像向量进行降维并提取其全局特征,对降维后的热红外人脸全局特征采用LDA算法训练生成一个使类间离散度最大、类内离散度最小的最佳分类器.最后,进行基于PCA+LDA的热红外人脸图像识别研究,实验结果表明该方法可获得较高的识别率. 相似文献
4.
为了提高人脸识别效率,提出了一种基于PCA、LDA和SVM算法融合的人脸识别方法。使用主成分分析(PCA)将人脸图像变换到新的特征空间中,消除图像特征间的相关性和噪声,提取人脸全局特征,在实验阶段取较多的投影方向使其尽可能多的保持原始信息;使用线性判别分析(LDA)算法进一步投影变换降低数据维度;使用支持向量机(SVM)分类识别。将PCA、LDA和SVM三种算法的优点结合起来,在ORL数据库上进行仿真实验,结果表明该方法的识别率可达99.0%。 相似文献
5.
本文首先介绍了LDA方法在人脸识别上的应用,然后介绍了直接用LDA方法会遇到的问题,提出了采用先降维再使用LDA方法的思想。并且进行了识别率的比较。 相似文献
6.
在2维线性鉴别分析(2DLDA)的基础上.介绍了2维异方差鉴别分析(2DHDA),并将其应用于人脸识别.2DHDA算法去除了2DLDA算法样本类内协方差相等的约束,克服了异方差鉴别分析(HDA)算法的"小样本"问题.首先,根据2DLDA准则定义2DHDA准则;然后,将2DHDA准则取对数,用梯度下降法求得最优投影矩阵,人脸图像向最优投影矩阵投影提取人脸图像的特征;最后,最小距离分类器完成人脸识别.基于ORL与Yale混合人脸数据库的实验结果表明了2DHDA应用于人脸识别的有效性. 相似文献
7.
利用相似度多个维度的信息进行开集判别,以提高开集人脸识别的准确率。该方法首先通过大量带标识的测试样本获得已知类样本和非已知类样本相似度向量的分布,然后引入线性判别分析学习两个类中相似度向量的分布特征,在开集判别中通过相似度向量的特征匹配来判断样本是否为已知类。利用相似度分布中的分类信息,训练出的特征具有更强的分类能力。不同人脸库的实验表明,相对于传统方法,文中方法能提高开集识别的准确率。 相似文献
8.
提出了二维主成分分析(2DPCA)与二维线性鉴别分析(2DLDA)相结合的双向压缩投影的子空间人脸识别方法.该方法在进行一次2DPCA运算后,对特征矩阵进行转置,再进行2DLDA运算,与(2D)~2PCA与(2D)~2LDA相比,充分利用了2DPCA和2DLDA的优点,既包含了样本的类别信息,又消除了图像矩阵行和列的相关性,有效地提取了行和列的识别信息,识别特征维数也大幅度减少.在ORL和PERET人脸库上的实验表明,在不影响识别速度的情况下,其识别率优于现有二维特征提取方法. 相似文献
9.
10.
11.
12.
人脸识别中线性判别分析的单参数正则化方法 总被引:1,自引:1,他引:1
将线性判别分析(LDA)应用于人脸识别中时,小样本问题常常出现,即,通常可获得的人脸训练样本个数远小于训练样本的维数,从而导致类内散布矩阵Sw奇异,于是得到病态的特征值问题.使用数学工具探讨了这一现象的实质.此外,提出了一种单参数正则化方法来解决小样本问题,该方法以满足tr(S'w)=tr(Sw)为条件,用一个可逆矩阵S'w去估计奇异的类内散布矩阵Sw.在使用小波变换对人脸像降维预处理后进行了该方法与传统LDA的对比实验.实验表明,该方法可大幅提高LDA的识别性能. 相似文献
13.
14.
15.
NNSRM is an implementation of the structural risk minimization (SRM) principle using the nearest neighbor (NN) rule, and linear
discriminant analysis (LDA) is a dimension-reducing method, which is usually used in classifications. This paper combines
the two methods for face recognition. We first project the face images into a PCA subspace, then project the results into
a much lower-dimensional LDA subspace, and then use an NNSRM classifier to recognize them in the LDA subspace. Experimental
results demonstrate that the combined method can achieve a better performance than NN by selecting different distances and
a comparable performance with SVM but costing less computational time.
Danian Zheng received his Bachelor degree in Computer Science and Technology in 2002 from Tsinghua University, Beijing, China. He received his Master degree and Doctoral degree in Computer Science and Technology in 2006 from Tsinghua University. He is currently a researcher in Fujitsu R&D Center Co. Ltd, Beijing, China. His research interests are mainly in the areas of support vector machines, kernel methods and their applications. Meng Na received her Bachelor degree in Computer Science and Technology in 2003 from Northeastern, China. Since 2003 she has been pursuing the Master degree and the Doctoral degree at the Department of Computer Science and Technology at Tsinghua University. Her research interests are in the area of image processing, pattern recognition, and virtual human. Jiaxin Wang received his Bachelor degree in Automatic Control in 1965 from Beijing University of Aeronautics and Astronautics, his Master degree in Computer Science and Technology in 1981 from Tsinghua University, Beijing, China, and his Doctoral degree in 1996 from Engineering Faculty of Vrije Universiteit Brussel, Belgium. He is currently a professor of Department of Computer Science and Technology, Tsinghua University. His research interests are in the areas of artificial intelligence, intelligent control and robotics, machine learning, pattern recognition, image processing and virtual reality. 相似文献
Jiaxin Wang (Corresponding author)Email: |
Danian Zheng received his Bachelor degree in Computer Science and Technology in 2002 from Tsinghua University, Beijing, China. He received his Master degree and Doctoral degree in Computer Science and Technology in 2006 from Tsinghua University. He is currently a researcher in Fujitsu R&D Center Co. Ltd, Beijing, China. His research interests are mainly in the areas of support vector machines, kernel methods and their applications. Meng Na received her Bachelor degree in Computer Science and Technology in 2003 from Northeastern, China. Since 2003 she has been pursuing the Master degree and the Doctoral degree at the Department of Computer Science and Technology at Tsinghua University. Her research interests are in the area of image processing, pattern recognition, and virtual human. Jiaxin Wang received his Bachelor degree in Automatic Control in 1965 from Beijing University of Aeronautics and Astronautics, his Master degree in Computer Science and Technology in 1981 from Tsinghua University, Beijing, China, and his Doctoral degree in 1996 from Engineering Faculty of Vrije Universiteit Brussel, Belgium. He is currently a professor of Department of Computer Science and Technology, Tsinghua University. His research interests are in the areas of artificial intelligence, intelligent control and robotics, machine learning, pattern recognition, image processing and virtual reality. 相似文献
16.
小波变换后的低频子带图像既去除了某些表情变化,又减小了数据量,而图像的频谱特征则具有良好分类特性,因此两者结合后得到的频谱脸在人脸识别方面具有相当高的应用价值。先利用小波变换和Fourier变换求得原始人脸图像的频谱脸(Spectrofaces),再对频谱脸继续求取各自的本征脸(Eigenface)和LDA(Linear Discriminant Analysis)特征作为分类特征,并利用了不同的分类方法进行识别。实验是利用ORL人脸库进行的,实验结果证明了比起直接利用空间域上原始图像的识别方法来说,基于频谱的方法可以有效提高识别率。 相似文献
17.
自动人脸识别技术综述 总被引:5,自引:2,他引:5
1.引言人脸识别因其在安全、认证、人机交互等方面良好的应用前景而成为近年来模式识别和人工智能领域的研究热点。到目前为止,基于计算机的自动人脸识别系统已经研究了20多年,但仍面临着种种困难,这些困难来源于表情、姿态、位置、头部尺寸、光照和背景的大幅度变化。同时,由于 相似文献
18.
结合整体与局部信息的人脸识别方法 总被引:3,自引:0,他引:3
提出了一种综合利用整体和局部信息进行人脸识别的新方法。在对整幅人脸图像进行PCA分析的基础上辅以了局部区域的PCA LDA分析。在ORL的400幅人脸库上对此方法进行了验证,结果证明此方法是有效可行的,最优的识别率达到了97%,比仅利用完整图像的人脸识别方法有了不同程度的提高。 相似文献