CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleResearch on Fusion Approaches of Face Recognition with different Imagery
Thesis Advisor刘昌平
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword图像处理 融合 人脸识别 异构图像 近红外 Image Processing Fusion Face Recognition Different Imagery Infrared
Abstract光照条件的变化是导致人脸识别率下降的主要原因之一。当人在室内进行人脸注册,室内条件下可以正常识别, 但在室外的识别效果就非常差。 这就需要针对实际工作中的人脸识别问题提出一种方法,这种方法首先要能够解决在光照环境极端条件下的人脸识别问题, 但同时不能影响正常光照条件下的识别。 本文提出两种异构人脸图像融合:原始图像和光照预处理后人脸图像融合,可见光和近红外光人脸图像融合,分别基于分值匹配层和 决策层的融合。在分值层的融合上提出了基于SRC(Sparse Representation Classify)分类器的融合方法以及基于最近邻分类器的融合方法。在决策层的融合提出了基于投票分类的改进融合方法。 本文提出的融合方法如下: (1)在分值匹配层提出了基于SRC分类器的融合方法。首先,对人脸进行预处理,并使用处理后的人脸图像集和原 始人脸图像集作为训练样本。其次,针对扩充后的两个训练样本集--原始人脸图像和预处理变换后的图像,在SRC 分类器基础上,提出了一种融合方法用于提升识别性能。 (2)在分值匹配层同时提出了基于最近邻分类器的融合方法。对于原始图像,与原始图像注册集进行比对, 挑选出其中最匹配的两个模板并计算相应的相似度,同样对于预处理图像,与预处理图像注册集进行比对, 挑选出其中最匹配的两个处理后的模板并计算相应的相似度。最后通过这四个相似度计算出对应的两个相对分数值, 并通过比较分数值的大小判断采用哪种图像的识别结果。 (3)在决策层提出了基于投票分类的改进融合方法,该方法引进了分值参考,共可以分为四个步骤:类别排名,分值归一化,分值加和及排序。
Other AbstractThe effect of variation in the illumination conditions, which causes dramatic changes in the face appearance, is one of the challenging problems in face recognition. For more robust face recognition, the paper proposed two methods to deal with the situation in varying lighting, whether the query data and the corresponding gallery data are acquired in the same conditions or in the different illumination conditions. The one is based on processed and raw facial images, which fused in the score level, and the other is based on visual and near-infrared facial images, which fused in the decision level. The work of the paper includes the following aspects: (1) In the score level, the fusion method is based on the ~SRC(Sparse Representation Classify).The method added raw grayscale images to the gallery and query sets, and got two distance measures or similarity measures, also known as scores, through ~SRC respectively. Finally, the method combined the two scores to get the final result. (2) In the score level, the second fusion method is based on the NN(Nearest Neighbor) Classify. Firstly, we define a score to denote a relative difference of the first and second largest similarities between the query input and the individuals in the gallery classes. Then, according to the score, we choose the appropriate images, raw or processed images, to involve the recognition. (3) In the decision level, the fusion method is based on the vote counting. It consists of rank list of classes,normalization, score adding and score ranking.
Other Identifier200828014629085
Document Type学位论文
Recommended Citation
GB/T 7714
许力. 不同光照下的异构人脸图像的融合识别方法[D]. 中国科学院自动化研究所. 中国科学院研究生院,2008.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20082801462908(2763KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[许力]'s Articles
Baidu academic
Similar articles in Baidu academic
[许力]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[许力]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.