卷积鲁棒主成分分析

鲁棒主成分分析(RPCA)是一种经典的高维数据分析方法,可从带噪声的观测样本中恢复出原始数据.但是,RPCA能工作的前提是目标数据拥有低秩矩阵结构,不能有效处理实际应用中广泛存在的非低秩数据.研究发现,虽然图像、视频等数据矩阵本身可能不是低秩的,但它们的卷积矩阵通常是低秩的.根据这一原理,提出一种称为卷积鲁棒主成分分析...     

A robust approach to independent component analysis of signals with high-level noise measurements

We propose a robust approach for independent component analysis (ICA) of signals where observations are contaminated with high-level additive noise and/or outliers. The source signals may contain mixtures of both sub-Gaussian and super-Gaussian components, and the number of sources is unknown. Our robust approach includes two procedures. In the first procedure, a robust prewhitening technique is used to reduce the power of additive noise, the dimensionality and the correlation among sources. A cross-validation technique is introduced to estimate the number of sources in this first procedure. In the second procedure, a nonlinear function is derived using the parameterized t-distribution density model. This nonlinear function is robust against the undue influence of outliers fundamentally. Moreover, the stability of the proposed algorithm and the robust property of misestimating the parameters (kurtosis) have been studied. By combining the t-distribution model with a family of light-tailed distributions (sub-Gaussian) model, we can separate the mixture of sub-Gaussian and super-Gaussian source components. Through the analysis of artificially synthesized data and real-world magnetoencephalographic (MEG) data, we illustrate the efficacy of this robust approach.     

基于多线性主元分析和FLD的掌纹识别*

为了减少高维对计算成本的影响,同时提取有利于分类的判别特征,提出运用多线性主元分析(MPCA)与FLD相结合的方法进行掌纹识别。运用MPCA直接对掌纹张量进行降维和特征提取,低维特征向量作为FLD的输入,提取判别特征向量,计算特征向量间的余弦距离进行掌纹匹配。PolyU掌纹图像库的实验结果表明,与主元分析（PCA）、PCA+FLD、二维主元分析(2DPCA)、独立元分析（ICA）和MPCA相比,该算法的识别率(RR)最高为9991%,特征提取和匹配总时间为0398 s,满足实时系统的要求。     

不完全鲁棒主成分分析的正则化方法及其在背景建模中的应用

针对现有的鲁棒主成分分析(RPCA)方法忽略序列数据的连续性及不完整性的情况,提出了一种低秩矩阵恢复模型——正则化不完全鲁棒主成分分析(RIRPCA)。首先基于序列数据连续性的度量函数建立了RIRPCA模型,即最小化矩阵核范数、L₁范数和正则项的加权组合;然后使用增广拉格朗日乘子法来求解所提出的凸优化模型, 此算法具有良好的可扩展性和较低的计算复杂度;最后,将RIRPCA应用到视频背景建模中。实验结果表明,RIRPCA比矩阵补全和不完全RPCA等方法在恢复丢失元素和分离前景上具有优越性。     

Visual learning and recognition of 3D objects using two-dimensional principal component analysis: A robust and an efficient approach

Inspired by the conviction that the successful model employed for face recognition [M. Turk, A. Pentland, Eigenfaces for recognition, J. Cogn. Neurosci. 3(1) (1991) 71-86] should be extendable for object recognition [H. Murase, S.K. Nayar, Visual learning and recognition of 3-D objects from appearance, International J. Comput. Vis. 14(1) (1995) 5-24], in this paper, a new technique called two-dimensional principal component analysis (2D-PCA) [J. Yang et al., Two-dimensional PCA: a new approach to appearance based face representation and recognition, IEEE Trans. Patt. Anal. Mach. Intell. 26(1) (2004) 131-137] is explored for 3D object representation and recognition. 2D-PCA is based on 2D image matrices rather than 1D vectors so that the image matrix need not be transformed into a vector prior to feature extraction. Image covariance matrix is directly computed using the original image matrices, and its eigenvectors are derived for feature extraction. The experimental results indicate that the 2D-PCA is computationally more efficient than conventional PCA (1D-PCA) [H. Murase, S.K. Nayar, Visual learning and recognition of 3-D objects from appearance, International J. Comput. Vis. 14(1) (1995) 5-24]. It is also revealed through experimentation that the proposed method is more robust to noise and occlusion.     

Laplacian regularized robust principal component analysis for process monitoring

Principal component analysis (PCA) is one of the most widely used techniques for process monitoring. However, it is highly sensitive to sparse errors because of the assumption that data only contains an underlying low-rank structure. To improve classical PCA in this regard, a novel Laplacian regularized robust principal component analysis (LRPCA) framework is proposed, where the “robust” comes from the introduction of a sparse term. By taking advantage of the hypergraph Laplacian, LRPCA not only can represent the global low-dimensional structures, but also capture the intrinsic non-linear geometric information. An efficient alternating direction method of multipliers is designed with convergence guarantee. The resulting subproblems either have closed-form solutions or can be solved by fast solvers. Numerical experiments, including a simulation example and the Tennessee Eastman process, are conducted to illustrate the improved process monitoring performance of the proposed LRPCA.     

鲁棒主元分析在掌纹识别中的应用

为了对存在异常值的图像构建低维线性子空间的描述,提出用鲁棒主元分析(RPCA)的新方法进行掌纹识别。运用图像下抽样方法降低掌纹空间的维数,在低维图像上应用RPCA提取低维的投影向量,然后将训练图像和待识别图像向投影向量上投影得到鲁棒主元特征,计算特征向量间的余弦距离进行掌纹匹配。运用PolyU掌纹图像库进行测试,结果表明,与主元分析(PCA)、独立元分析(ICA)和核主元分析(KPCA)相比,RPCA算法的识别率最高为99%,特征提取和匹配总时间0.032 s,满足了实时系统的要求。     

A new multidimensional discriminant representation for robust person re-identification

Pattern Analysis and Applications - Person Re-Identification (PRe-ID) or person retrieval is a challenging task of computer vision, aiming to identify a specific person across disjoint cameras...     

牛顿-软阈值迭代鲁棒主成分分析算法

针对鲁棒主成分分析（RPCA）问题,为了降低RPCA算法的时间复杂度,提出了牛顿-软阈值迭代（NSTI）算法。首先,使用低秩矩阵的Frobenius范数与稀疏矩阵的l₁-范数的和来构造NSTI算法的模型;其次,同时使用两种不同的优化方式求解模型的不同部分,即用牛顿法快速计算出低秩矩阵,用软阈值迭代算法快速计算出稀疏矩阵,交替使用这两种方法计算出原数据的低秩矩阵和稀疏矩阵的分解;最后,得到原始数据的低秩特征。在数据规模为5 000×5 000,低秩矩阵的秩为20的情况下,NSTI算法和梯度下降（GD）算法、低秩矩阵拟合（LMaFit）算法相比,时间效率分别提高了24.6%、45.5%。对180帧的视频前景背景进行分离,NSTI耗时3.63 s,时间效率比GD算法、LMaFit算法分别高78.7%、82.1%。图像降噪实验中,NSTI算法耗时0.244 s,所得到的降噪后的图像与原始图像的残差为0.381 3,与GD算法、LMaFit算法相比,时间效率和精确度分别提高了64.3%和45.3%。实验结果证明,NSTI算法能够有效解决RPCA问题并提升RPCA算法的时间效率。     

Non-linear feature extraction for robust speech recognition in stationary and non-stationary noise

An analysis-based non-linear feature extraction approach is proposed, inspired by a model of how speech amplitude spectra are affected by additive noise. Acoustic features are extracted based on the noise-robust parts of speech spectra without losing discriminative information. Two non-linear processing methods, harmonic demodulation and spectral peak-to-valley ratio locking, are designed to minimize mismatch between clean and noisy speech features. A previously studied method, peak isolation [IEEE Transactions on Speech and Audio Processing 5 (1997) 451], is also discussed with this model. These methods do not require noise estimation and are effective in dealing with both stationary and non-stationary noise. In the presence of additive noise, ASR experiments show that using these techniques in the computation of MFCCs improves recognition performance greatly. For the TI46 isolated digits database, the average recognition rate across several SNRs is improved from 60% (using unmodified MFCCs) to 95% (using the proposed techniques) with additive speech-shaped noise. For the Aurora 2 connected digit-string database, the average recognition rate across different noise types, including non-stationary noise background, and SNRs improves from 58% to 80%.     

An iterative algorithm for robust kernel principal component analysis

We introduce a technique to improve iterative kernel principal component analysis (KPCA) robust to outliers due to undesirable artifacts such as noises, alignment errors, or occlusion. The proposed iterative robust KPCA (rKPCA) links the iterative updating and robust estimation of principal directions. It inherits good properties from these two ideas for reducing the time complexity, space complexity, and the influence of these outliers on estimating the principal directions. In the asymptotic stability analysis, we also show that our iterative rKPCA converges to the weighted kernel principal kernel components from the batch rKPCA. Experimental results are presented to confirm that our iterative rKPCA achieves the robustness as well as time saving better than batch KPCA.     

牛顿-软阈值迭代鲁棒主成分分析算法

针对鲁棒主成分分析（RPCA）问题，为了降低RPCA算法的时间复杂度，提出了牛顿-软阈值迭代（NSTI）算法。首先，使用低秩矩阵的Frobenius范数与稀疏矩阵的l₁-范数的和来构造NSTI算法的模型；其次，同时使用两种不同的优化方式求解模型的不同部分，即用牛顿法快速计算出低秩矩阵，用软阈值迭代算法快速计算出稀疏矩阵，交替使用这两种方法计算出原数据的低秩矩阵和稀疏矩阵的分解；最后，得到原始数据的低秩特征。在数据规模为5 000×5 000，低秩矩阵的秩为20的情况下，NSTI算法和梯度下降（GD）算法、低秩矩阵拟合（LMaFit）算法相比，时间效率分别提高了24.6%、45.5%。对180帧的视频前景背景进行分离，NSTI耗时3.63 s，时间效率比GD算法、LMaFit算法分别高78.7%、82.1%。图像降噪实验中，NSTI算法耗时0.244 s，所得到的降噪后的图像与原始图像的残差为0.381 3，与GD算法、LMaFit算法相比，时间效率和精确度分别提高了64.3%和45.3%。实验结果证明，NSTI算法能够有效解决RPCA问题并提升RPCA算法的时间效率。     

结合多尺度时频调制与多线性主成分分析的乐器识别

针对目前时域频域特征、倒谱特征、稀疏特征、概率特征对同族乐器错分率高且对打击乐器识别不佳的问题,提出一种提取时频信息且低冗余度的模型用于乐器识别。首先利用耳蜗模型对乐音进行谐波分解,生成接近人耳感知且包含时频信息的听觉谱图（AS）;随后利用多尺度滤波器对听觉谱图多尺度时频调制（MTFM）以观测时频的变化;最后利用多线性主成分分析（MPCA）对调制输出在保留数据内在相关的前提下降维,并使用支持向量机（SVM）分类。仿真实验表明,该方法在IOWA数据库上取得92.74%的正确率,对打击乐器与同族乐器的错分率分别为3%与9.12%,均优于上述特征。相比主成分分析（PCA）降维,MPCA提高识别准确率6.43%。因此,该模型适用于对同族乐器与打击乐器的识别。     

A new subspace identification approach based on principal component analysis

Principal component analysis (PCA) has been widely used for monitoring complex industrial processes with multiple variables and diagnosing process and sensor faults. The objective of this paper is to develop a new subspace identification algorithm that gives consistent model estimates under the errors-in-variables (EIV) situation. In this paper, we propose a new subspace identification approach using principal component analysis. PCA naturally falls into the category of EIV formulation, which resembles total least squares and allows for errors in both process input and output. We propose to use PCA to determine the system observability subspace, the A, B, C, and D matrices and the system order for an EIV formulation. Standard PCA is modified with instrumental variables in order to achieve consistent estimates of the system matrices. The proposed subspace identification method is demonstrated using a simulated process and a real industrial process for model identification and order determination. For comparison the MOESP algorithm and N4SID algorithm are used as benchmarks to demonstrate the advantages of the proposed PCA based subspace model identification (SMI) algorithm.     

鲁棒主成分分析的运动目标检测综述

目的 运动目标检测在许多计算机视觉任务中发挥了重要的作用。背景建模是运动目标检测中传统而又常用的方法。然而,许多背景建模方法是基于像素点的,对背景方面的考虑过于简单,难于处理真实视频。最近,将基于低秩和稀疏分解的鲁棒主成分分析应用于运动目标检测成为计算机视觉领域内的研究热点。为使更多国内外运动目标检测的研究者对鲁棒主成分分析方法进行探索和应用,本文对其进行系统综述。方法 融入最新研究进展,基于误差抑制、贝叶斯理论、时间和空间信息、多特征和多因素耦合,对各种国内外的鲁棒主成分分析模型进行归纳,并理论分析其优缺点。结果 本文采用变化检测数据集（change detection dataset）中不同场景的视频序列来对不同算法进行对比实验。从实验结果可知,属于第3类方法的DECOLOR 的检测效果优于其他算法,在均值对比中得到的召回率、精确率和F-measure分别为0.7、0.706和0.66。总体来说,当前改进算法都能有效地弥补最初鲁棒主成分分析方法的缺陷,提高了运动目标检测的精度。结论 鲁棒主成分分析在运动目标检测上取得了较多的研究与应用成果,在智能视频监控应用领域拥有广阔的应用前景。但是,其仍需针对鲁棒主成分分析存在的一些局限性进行深入的研究。融入前景运动目标在视频中的先验知识是基于鲁棒主成分分析的运动目标检测的发展趋势。     

基于鲁棒主成分分析的Canny边缘检测算法

为提高图像边缘检测的准确性和鲁棒性,提出一种基于鲁棒主成分分析(RPCA)的Canny边缘检测算法。该算法对图像进行RPCA分解得到图像的主成分和稀疏成分,利用Canny算子对主成分进行边缘检测,从而实现对图像的边缘检测。该算法将图像的边缘检测问题转化为图像主成分的边缘检测问题,消除了图像信息中“污点”对检测结果的干扰,抑制了噪声。仿真实验结果表明,该算法在边缘检测的准确性和鲁棒性方面优于Log边缘检测算法、Canny边缘检测算法和Susan边缘检测算法方法。     

多特征稳健主成分分析的视频运动目标分割

提出一种多特征稳健主成分分析(MFRPCA)算法,该算法融合多种视觉特征进行视频运动目标分割,分割的目的即将运动目标从静止信息中提取出来,分割的主要过程是将多特征视频矩阵分解为低秩矩阵和稀疏矩阵.矩阵分解过程是求解一个带受限条件的核范数与L2,1范数组合的最小化问题,此最小化问题可以通过增广拉格朗日乘子法(ALM)有效求解.与其他算法相比,本文算法融合了图像的颜色、边缘和纹理特征等多个特征,通过对变化检测基准数据集进行检测,本文算法获得的查全率为0.486 0和F度量为0.559 7,实验结果表明,本文算法的稳健性和可靠性均优于其他算法.     

A fast pixel mapping algorithm using principal component analysis

Pixel mapping is one of the basic processes in color quantization. In this paper, we shall propose a new algorithm using principal component analysis as a faster approach to pixel mapping. Within much shorter search time, our new scheme can find the nearest color which is identical to the one found using a full search. The idea behind the proposed method is quite simple. First, we compute two principal component directions (PCDs) for the palette. Then, the projected values on PCDs are computed for each color in palette. Finally, the projected values, following the triangular inequality principle, can help us reduce the computation time for finding the nearest color. The experimental results reveal that the proposed scheme is more efficient than the previous work.     

Improved principal component monitoring using the local approach

This paper shows that current multivariate statistical monitoring technology may not detect incipient changes in the variable covariance structure nor changes in the geometry of the underlying variable decomposition. To overcome these deficiencies, the local approach is incorporated into the multivariate statistical monitoring framework to define two new univariate statistics for fault detection. Fault isolation is achieved by constructing a fault diagnosis chart which reveals changes in the covariance structure resulting from the presence of a fault. A theoretical analysis is presented and the proposed monitoring approach is exemplified using application studies involving recorded data from two complex industrial processes.     

Unsupervised band extraction for hyperspectral images using clustering and kernel principal component analysis

The ‘curse of dimensionality’ is a drawback for classification of hyperspectral images. Band extraction is a technique for reducing the dimensionality and makes it computationally less complex for classification. In this article, an unsupervised band extraction method for hyperspectral images has been proposed. In the proposed method, kernel principal component analysis (KPCA) is used for transformation of the original data, which integrates the nonlinear characteristics, as well as, the advantages of principal component analysis and extract higher order statistics of data. The KPCA is highly dependent on the number of patterns for calculating kernel matrix. So, a proper selection of subset of patterns, which represent the original data properly, may reduce the computational cost for the proposed method with considerably better performance. Here, density-based spatial clustering technique is first used to group the pixels according to their similarity, and then some percentages of pixels from each cluster are selected to form the proper subset of patterns. To demonstrate the effectiveness of the proposed clustering- and KPCA-based unsupervised band extraction method, investigation is carried out on three hyperspectral data sets, namely, Indian, KSC, and Botswana. Four evaluation measures, namely classification accuracy, kappa coefficient, class separability, and entropy are calculated over the extracted bands to measure the efficiency of the proposed method. The performance of the proposed method is compared with four state-of-the-art unsupervised band extraction approaches, both qualitatively and quantitatively, and shows promising results compared to them in terms of four evaluation measures.