基于二维特征提取和稀疏表示的人脸识别算法研究
发布时间：2018-08-02 08:56
【摘要】：近几年出现的压缩感知理论突破了传统采样对于采样率的限制,为信号采集和信号处理提供了新的思路。压缩感知引起了很多学者的关注,根据压缩感知理论,有学者提出了一种新的人脸识别算法——稀疏表示分类器(SRC)。SRC算法的创新点就在于其直接利用训练集中的人脸图像构成基矩阵,而后将测试集中的人脸图片用基矩阵线性表示,求出最稀疏解,最后就可以从最稀疏解中提取出分类信息。但是如果把SRC算法直接应用到人脸图像上,算法的计算复杂度太高。在本文中我们提出了一种改进算法,就是把SRC算法和二维特征提取算法相结合。在利用SRC算法进行人脸识别之前,我们先用二维特征提取算法对原始图片进行特征提取和降维。这样不仅可以降低算法的计算复杂度,还可以保留图像的结构特征。本文的主要工作包括:首先比较了常见的压缩感知重构算法,并且选择了 OMP算法作为重构算法;其次改进现有的SRC算法,加入二维特征提取算法来降低算法的计算复杂度;最后在AR人脸数据库上进行实验仿真,来验证改进算法的可行性和优越性。通过对实验结果的分析对比,证明我们提出的改进算法是可行的,并且在识别率和识别花费的时间上是优于现有算法的。
[Abstract]:In recent years, the theory of compression sensing has broken through the limitation of traditional sampling rate and provided a new idea for signal acquisition and signal processing. Compression perception has attracted the attention of many scholars. According to the theory of compressed perception, Some scholars have proposed a new face recognition algorithm, the sparse representation classifier (SRC) SRC), whose innovation lies in its direct use of the face images in the training set to form the base matrix, and then the face images in the test set are expressed linearly by the basis matrix. Finally, the classification information can be extracted from the sparse solution. However, if the SRC algorithm is directly applied to face images, the computational complexity of the algorithm is too high. In this paper, we propose an improved algorithm, which combines SRC algorithm with two-dimensional feature extraction algorithm. Before using the SRC algorithm for face recognition, we use the two-dimensional feature extraction algorithm to extract and reduce the dimension of the original image. This not only reduces the computational complexity of the algorithm, but also preserves the structural features of the image. The main work of this paper is as follows: firstly, the common compression perception reconstruction algorithms are compared, and the OMP algorithm is selected as the reconstruction algorithm; secondly, the existing SRC algorithm is improved, and two-dimensional feature extraction algorithm is added to reduce the computational complexity of the algorithm. Finally, the experiment is carried out on AR face database to verify the feasibility and superiority of the improved algorithm. Through the analysis and comparison of the experimental results, it is proved that the proposed improved algorithm is feasible, and is superior to the existing algorithm in recognition rate and the time spent in recognition.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
[Abstract]:In recent years, the theory of compression sensing has broken through the limitation of traditional sampling rate and provided a new idea for signal acquisition and signal processing. Compression perception has attracted the attention of many scholars. According to the theory of compressed perception, Some scholars have proposed a new face recognition algorithm, the sparse representation classifier (SRC) SRC), whose innovation lies in its direct use of the face images in the training set to form the base matrix, and then the face images in the test set are expressed linearly by the basis matrix. Finally, the classification information can be extracted from the sparse solution. However, if the SRC algorithm is directly applied to face images, the computational complexity of the algorithm is too high. In this paper, we propose an improved algorithm, which combines SRC algorithm with two-dimensional feature extraction algorithm. Before using the SRC algorithm for face recognition, we use the two-dimensional feature extraction algorithm to extract and reduce the dimension of the original image. This not only reduces the computational complexity of the algorithm, but also preserves the structural features of the image. The main work of this paper is as follows: firstly, the common compression perception reconstruction algorithms are compared, and the OMP algorithm is selected as the reconstruction algorithm; secondly, the existing SRC algorithm is improved, and two-dimensional feature extraction algorithm is added to reduce the computational complexity of the algorithm. Finally, the experiment is carried out on AR face database to verify the feasibility and superiority of the improved algorithm. Through the analysis and comparison of the experimental results, it is proved that the proposed improved algorithm is feasible, and is superior to the existing algorithm in recognition rate and the time spent in recognition.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 王李冬;;一种新的人脸识别算法[J];计算机技术与发展;2009年05期
2 桂荣;沈荣鑫;詹泳;;基于模糊神经推理的人脸识别算法研究[J];华南师范大学学报(自然科学版);2012年04期
3 苏士美;王燕;王明霞;;基于加权小波分解的人脸识别算法研究[J];郑州大学学报(工学版);2014年01期
4 魏冬梅;周卫东;;采用压缩感知的人脸识别算法[J];计算机工程;2011年18期
5 陈皓;霍星;;视频监控中人脸识别算法稳定性的改进[J];工程图学学报;2011年06期
6 李美瑾;;人脸识别算法综述与远程考试系统应用研究[J];广西广播电视大学学报;2012年01期
7 盛朝强;王君;;煤矿井下人员签到系统人脸识别算法研究[J];计算机技术与发展;2012年07期
8 朱学毅;王崇骏;周新民;张W，
本文编号：2158909
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2158909.html
