计算机科学 ›› 2018, Vol. 45 ›› Issue (5): 238-242.doi: 10.11896/j.issn.1002-137X.2018.05.041
李小薪,吴克宋,齐盼盼,周旋,刘志勇
LI Xiao-xin, WU Ke-song, QI Pan-pan, ZHOU Xuan and LIU Zhi-yong
摘要: 当人脸图像中存在较大比例的光照变化或遮挡时,PCANet所采用的局部零均值预处理以及PCA滤波器对噪声的过滤作用将导致所生成的特征图的整体分布主要集中在0附近,在一定程度上丧失了冗余性。为了提升PCANet对抗噪声的能力,提出了局部球面规范化方法,并将其嵌入PCANet的卷积层,从而拓展了PCANet特征的丰富性。在UMBDB和AR库上的实验表明,改进后的PCANet具有更好的冗余性、鲁棒性和识别性能。一个重要的发现在于:特征的冗余性需要在噪声滤除的过程中逐步提升,直接对输入图像施加局部球面规范化可能会导致不稳定的识别性能。
[1] IOFFE S,SZEGEDY C.Batch Normalization:Accelerating Deep Network Training by Reducing Internal Covariate Shift [C]∥International Conference on Machine Learning.2015:448-456. [2] CHAN T,JIA K,GAO S,et al.PCANet:A Simple Deep Lear-ning Baseline for Image Classification? [J].IEEE Transactions on Image Processing,2015,24(12):5017-5032. [3] AHONEN T,HADID A,PIETIKAINEN M.Face Description with Local Binary Patterns:Application to Face Recognition [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(12):2037-2041. [4] PARKHI O M,VEDALDI A,ZISSERMAN A.Deep Face Reco-gnition [C]∥Proceedings of British Machine Vision Confe-rence.London:BMVA Press,2015:41. [5] WU X,HE R,SUN Z.A Lightened CNN for Deep Face Representation[J].arXiv.1511.02683,2015. [6] ZHANG L,YANG M,FENG X.Sparse Representation or Collaborative Representation:Which Helps Face Recognition?[C]∥IEEE International Conference on Computer Vision.2011:471-478. [7] COLOMBO A,CUSANO C,SCHETTINI R.UMB-DB:A Database of Partially Occluded 3D Faces [C]∥IEEE International Conference on Computer Vision Workshops.2011:2113-2119. [8] MARTINEZ A M.The AR Face Database[R].Columbus,USA:Computer Visual Center,Ohio State University,1998. |
No related articles found! |
|