自然科学版 英文版
自然科学版 英文版
自然科学版 英文版
自然科学版 英文版
英文版编委
自然科学版 英文版
英文版首届青年编委

您目前所在的位置:首页 - 期刊简介 - 详细页面

中南大学学报(自然科学版)

Journal of Central South University

第51卷    第2期    总第306期    2020年2月

[PDF全文下载]    [Flash在线阅读]

    

文章编号:1672-7207(2020)02-0377-08
基于多方向Gabor特征图协同表示的鲁棒人脸识别
张培1, 2,徐望明1, 2, 3,伍世虔1, 2,靳晓缘1, 2

(1. 武汉科技大学 信息科学与工程学院,湖北 武汉,430081;
2. 武汉科技大学 机器人与智能系统研究院,湖北 武汉,430081;
3. 武汉科技大学 教育部冶金自动化与检测技术工程研究中心,湖北 武汉,430081
)

摘 要: 为提高基于稀疏表示分类(SRC)算法在可变光照、姿态和表情下的人脸识别性能,提出一种基于多方向Gabor特征图(MGFM)和协同表示分类(CRC)的鲁棒人脸识别方法。首先,对人脸图像进行多方向多尺度Gabor变换,并融合同一方向不同尺度的Gabor特征;其次,在每个方向的融合特征图上提取Gist特征。在进行人脸识别时,可采取2种方法:1) 将人脸图像所有方向的Gist特征直接串联或自适应加权后串联构成人脸全局特征向量,并使用协同表示分类器得到识别结果;2) 对人脸图像每个方向的Gist特征向量分别使用协同表示分类器进行预分类,预分类时使用自适应K近邻策略确定候选类并进行评分,取总得分最高的类作为识别结果。最后,在ORL,Extended Yale B和AR等人脸数据库上开展人脸识别实验,由提出的方法分别取得99.8%,100%和99.7%的识别准确率和较快的执行速度。研究结果表明:本文方法利用多方向Gabor特征图(MGFM)建立人脸图像的特征表示能有效描述人脸局部信息,利用自适应K近邻策略改进协同表示分类算法能取得较高的识别准确率和执行效率。

 

关键词: 人脸识别;协同表示;多方向Gabor特征图;自适应K近邻

Robust face recognition based on collaborative representation of multi-directional Gabor feature maps
ZHANG Pei1, 2, XU Wangming1, 2, 3, WU Shiqian1, 2, JIN Xiaoyuan1, 2

1. School of Information Science and Engineering, Wuhan University of Science and Technology,Wuhan 430081, China;
2. Institute of Robotics and Intelligent Systems, Wuhan University of Science and Technology,Wuhan 430081, China;
3. Engineering Research Center for Metallurgical Automation and Detecting Technology of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081, China

Abstract:To improve the performance of face recognition based on SRC under conditions with varied illumination, pose and expression, a robust face recognition method based on multi-directional Gabor feature maps (MGFM) and collaborative representation classification(CRC) was proposed. Firstly, the multi-directional and multi-scale Gabor transform were performed on the face image, and the obtained Gabor features with different scales in the same direction were fused. Secondly, the Gist features were extracted for the fused feature maps in each direction. There were two ways which could be adopted to implement face recognition: 1) the Gist features of all directional feature maps of a face image were cascaded without or with adaptively-weighting to form the global feature vector of the face image and the recognition result was obtained based on collaborative representation classifier; 2) the pre-classification results were obtained by combining Gist features in each direction of a face image with collaborative representation classifiers respectively. The scores of the candidate classes were determined using the adaptive K nearest neighbor strategy, and the final recognition result had the highest total score Thirdly, experiments of face recognition were carried out on ORL, Extended Yale B and AR face database, and the proposed method reached the recognition accuracy of 99.8%, 100% and 99.7% respectively and obtained a faster execution speed. The results show that the proposed method can effectively describe the local information of face image using the multi-directional Gabor feature maps(MGFM), and the improved collaborative representation classification algorithm with the adaptive k-nearest neighbor strategy ultimately achieves higher recognition accuracy and execution efficiency.

 

Key words: face recognition; collaborative representation; multi-directional Gabor feature maps; adaptive K nearest neighbor

中南大学学报(自然科学版)
  ISSN 1672-7207
CN 43-1426/N
ZDXZAC
中南大学学报(英文版)
  ISSN 2095-2899
CN 43-1516/TB
JCSTFT
版权所有:《中南大学学报(自然科学版、英文版)》编辑部
地 址:湖南省长沙市中南大学 邮编: 410083
电 话: 0731-88879765(中) 88836963(英) 传真: 0731-88877727
电子邮箱:zngdxb@csu.edu.cn 湘ICP备09001153号