利用线性回归分类算法的人脸识别方法
|
摘要
针对人脸识别中存在的连续遮挡问题,笔者提出了一种利用线性回归分类算法的人脸识别新方法。首先,开发了一个线性模型,表示探针图像为特定类图库的一个线性组合。然后,对所有类模型的给定进行了探究,并且该决策是以有利于类的最小重建误差为规则。最后,对于连续遮挡问题,提出了一个模块化线性回归分类(LRC)方法进行分类识别,提出的LRC算法落入最近子空间分类。在人脸识别文献中的一些典型评估协议中,该算法在ORL人脸数据库上与集中先进算法进行评估。实验结果证明,提出的方法取得了98.75%的最高识别成功率。
Aiming at the continuous occlusion problem in face recognition, a new face recognition method based on linear regression classification algorithm is proposed. First, a linear model was developed to represent the probe image as a linear combination of a particular class library. Then a given probe for all class models, and the decision is based on a minimum reconstruction error that favors the class. Finally, for the continuous occlusion problem, a modular linear regression classification(LRC) method is proposed for classification and recognition, and the proposed LRC algorithm falls into the nearest subspace classification. In some typical evaluation protocols in the face recognition literature, the algorithm is evaluated on the ORL face database with a centralized advanced algorithm. The experimental results show that the proposed method achieves the highest recognition success rate of 98.75%.
Aiming at the continuous occlusion problem in face recognition, a new face recognition method based on linear regression classification algorithm is proposed. First, a linear model was developed to represent the probe image as a linear combination of a particular class library. Then a given probe for all class models, and the decision is based on a minimum reconstruction error that favors the class. Finally, for the continuous occlusion problem, a modular linear regression classification(LRC) method is proposed for classification and recognition, and the proposed LRC algorithm falls into the nearest subspace classification. In some typical evaluation protocols in the face recognition literature, the algorithm is evaluated on the ORL face database with a centralized advanced algorithm. The experimental results show that the proposed method achieves the highest recognition success rate of 98.75%.
引文
[1]刘君,黄燕琪,熊邦书.融合核主成分分析和最小距离鉴别投影的人脸识别算法[J].计算机工程,2016,42(4):221-225.
[2]成亚玲,谭爱平,张敏.混合多距离图像的线性判别分析人脸识别算法[J].系统仿真学报,2016,28(9):2254-2259.
[3]巫红霞,顾红其.基于曲波变换和独立成分分析的人脸识别[J].控制工程,2017,24(11):2319-2323.
[4]Zheng W,Zhao L,Zou C.A Modified Algorithm for Generalized Discriminant Analysis[J].Neural Computation,2004,16(6):1283-1297.
[5]基于稀疏与非线性表达的人脸识别问题研究[D].杭州:杭州电子科技大学,2016:78.
[6]周水生,郑颖,穆新亮.基于Cholesky分解的K2DPCA人脸识别研究[J].系统工程理论与实践,2016,36(2):528-535.
[7]Li C,Tan Y,Wang D,et al.Research on 3D Face Recognition Method in Cloud Environment Based on Semi Supervised Clustering Algorithm[J].Multimedia Tools&Applicatio ns,2017,76(16):17055-17073.
[8]Nouyed I,Poon B,Amin M A,et al.A Study on the Discriminating Characteristics of Gabor Phase-Face and an Improved Method For Face Recognition[J].International Journal of Machine Learning&Cybernetics,2016,7(6):1-16.
[2]成亚玲,谭爱平,张敏.混合多距离图像的线性判别分析人脸识别算法[J].系统仿真学报,2016,28(9):2254-2259.
[3]巫红霞,顾红其.基于曲波变换和独立成分分析的人脸识别[J].控制工程,2017,24(11):2319-2323.
[4]Zheng W,Zhao L,Zou C.A Modified Algorithm for Generalized Discriminant Analysis[J].Neural Computation,2004,16(6):1283-1297.
[5]基于稀疏与非线性表达的人脸识别问题研究[D].杭州:杭州电子科技大学,2016:78.
[6]周水生,郑颖,穆新亮.基于Cholesky分解的K2DPCA人脸识别研究[J].系统工程理论与实践,2016,36(2):528-535.
[7]Li C,Tan Y,Wang D,et al.Research on 3D Face Recognition Method in Cloud Environment Based on Semi Supervised Clustering Algorithm[J].Multimedia Tools&Applicatio ns,2017,76(16):17055-17073.
[8]Nouyed I,Poon B,Amin M A,et al.A Study on the Discriminating Characteristics of Gabor Phase-Face and an Improved Method For Face Recognition[J].International Journal of Machine Learning&Cybernetics,2016,7(6):1-16.