摘要
针对小波变换在嵌入式人脸识别中的不足,提出了Curvelet(曲波)变换结合2DPCA(二维主分量分析)的嵌入式人脸识别。采用Curvelet变换进行人脸图像特征的提取,经过比较,选取了Wrapping算法,然后利用2DPCA进行降维,并结合最近邻算法进行人脸识别。实验结果表明该方法很好得解决了人脸特征维数过高、数据量过大的缺点,识别效果更好,适用于嵌入式系统。
As Wavelet transform has disadvantages in face recognition on embedded system ,this paper proposes to use 2DPCA （two-dimensional principal component analysis） and Curvelet transform for face recognition on embedded system. We firstly use Curvelet transform to extract face image features with the Wrapping algorithm after comparing and then re- duce dimensions with the two-dimensional principal component analysis （2DPCA） ,at last the test face is recognized by using the nearest distance algorithm.The experiments demonstrate that the proposed method gets over the disadvantages of high dimension and large amount of data of face features,get a better recognition rate and is appropriate fur embedded system.
出处
《科技通报》
北大核心
2011年第5期750-753,共4页
Bulletin of Science and Technology
基金
江西师范大学科研项目(2009-7)
