Direct methods for recovering motion

We have developed direct methods for recovering the motion of an observer in a static environment in the case of pure rotation, pure translation, and arbitrary motion when the rotation is known. Some of these methods are based on the minimization of the difference between the observed time derivative of brightness and that predicted from the spatial brightness gradient, given the estimated motion. We minimize the square of the integral of this difference taken over the image region of interest. Other methods presented here exploit the fact that surfaces have to be in front of the observer in order to be seen.We do not establish point correspondences, nor do we estimate the optical flow. We use only first-order derivatives of the image brightness, and we do not assume an analytic form for the surface. We show that the field of view should be large to accurately recover the components of motion in the direction toward the image region. We also demonstrate the importance of points where the time derivative of brightness is small and discuss difficulties resulting from very large depth ranges. We emphasize the need for adequate filtering of the image data before sampling to avoid aliasing, in both the spatial and temporal dimensions.This research was supported by the National Science Foundation under Grant No. DMC85-11966. Additional support was provided by NASA (Grant No. GSFC 5-1162) and by the Veteran's Administration.BKPH on leave from the Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139.     

Subspace methods for face recognition

Studying the inherently high-dimensional nature of the data in a lower dimensional manifold has become common in recent years. This is generally known as dimensionality reduction. A very interesting strategy for dimensionality reduction is what is known as subspace analysis. Beginning with the Eigenface method, face recognition and in general computer vision has witnessed a growing interest in algorithms that capitalize on this idea and an ample number of such efficient algorithms have been proposed. These algorithms mainly differ in the kind of projection method used (linear or non-linear) or in the criterion employed for classification. The objective of this paper is to provide a comprehensive performance evaluation of about twenty five different subspace algorithms under several important real time test conditions. For this purpose, we have considered the performance of these algorithms on data taken from four standard face and object databases namely ORL, Yale, FERET and the COIL-20 object database. This paper also presents some theoretical aspects of the algorithm and the analysis of the simulations carried out.     

多代表点的子空间分类算法

多代表点近邻分类克服了传统近邻分类算法的缺点,使用以代表点为中心的模型簇构造分类模型并自动确定近邻数目.此类算法在不同类别的样本存在大量重叠时将导致模型簇数量增大,造成预测精度下降.提出了一种多代表点的子空间分类算法,将不同类别的训练样本投影到多个不同的子空间,使用子空间模型簇构造分类模型,有效分隔了不同类别样本在全空...     

Programmer-transparent coordination of recovering concurrentprocesses: philosophy and rules for efficient implementation

An approach to coordination of cooperating concurrent processes, each capable of error direction and recovery, is presented. Error detection, rollback, and retry in a process are specified by a well-structured language construct called recovery block. Recovery points of processes must be properly coordinated to prevent a disastrous avalanche of process rollbacks. The approach relies on an intelligent processor system (that runs processes) capable of establishing and discarding the recovery points of interacting processes in a well coordinated manner such that a process never makes two consecutive rollbacks without making a retry between the two, and every process rollback becomes a minimum-distance rollback. Following a discussion of the underlying philosophy of the author's approach, basic rules of reducing storage and time overhead in such a processor system are discussed. Examples are drawn from the systems in which processes communicate through monitors     

Enhanced Local Subspace Affinity for feature-based motion segmentation

We present a new motion segmentation algorithm: the Enhanced Local Subspace Affinity (ELSA). Unlike Local Subspace Affinity, ELSA is robust in a variety of conditions even without manual tuning of its parameters. This result is achieved thanks to two improvements. The first is a new model selection technique for the estimation of the trajectory matrix rank. The second is an estimation of the number of motions based on the analysis of the eigenvalue spectrum of the Symmetric Normalized Laplacian matrix. Results using the Hopkins155 database and synthetic sequences are presented and compared with state of the art techniques.     

自适应的软子空间聚类算法

软子空间聚类是高维数据分析的一种重要手段.现有算法通常需要用户事先设置一些全局的关键参数,且没有考虑子空间的优化.提出了一个新的软子空间聚类优化目标函数,在最小化子空间簇类的簇内紧凑度的同时,最大化每个簇类所在的投影子空间.通过推导得到一种新的局部特征加权方式,以此为基础提出一种自适应的k-means型软子空间聚类算法.该算法在聚类过程中根据数据集及其划分的信息,动态地计算最优的算法参数.在实际应用和合成数据集上的实验结果表明,该算法大幅度提高了聚类精度和聚类结果的稳定性.     

类属型数据核子空间聚类算法

现有的类属型数据子空间聚类方法大多基于特征间相互独立假设,未考虑属性间存在的线性或非线性相关性.提出一种类属型数据核子空间聚类方法.首先引入原作用于连续型数据的核函数将类属型数据投影到核空间,定义了核空间中特征加权的类属型数据相似性度量.其次,基于该度量推导了类属型数据核子空间聚类目标函数,并提出一种高效求解该目标函数的优化方法.最后,定义了一种类属型数据核子空间聚类算法.该算法不仅在非线性空间中考虑了属性间的关系,而且在聚类过程中赋予每个属性衡量其与簇类相关程度的特征权重,实现了类属型属性的嵌入式特征选择.还定义了一个聚类有效性指标,以评价类属型数据聚类结果的质量.在合成数据和实际数据集上的实验结果表明,与现有子空间聚类算法相比,核子空间聚类算法可以发掘类属型属性间的非线性关系,并有效提高了聚类结果的质量.     

一种基于数据流的软子空间聚类算法

针对高维数据的聚类研究表明,样本在不同数据簇往往与某些特定的数据特征子集相对应.因此,子空间聚类技术越来越受到关注.然而,现有的软子空间聚类算法都是基于批处理技术的聚类算法,不能很好地应用于高维数据流或大规模数据的聚类研究中.为此,利用模糊可扩展聚类框架,与熵加权软子空间聚类算法相结合,提出了一种有效的熵加权流数据软子空间聚类算法——EWSSC(entropy-weighting streaming subspace clustering).该算法不仅保留了传统软子空间聚类算法的特性,而且利用了模糊可扩展聚类策略,将软子空间聚类算法应用于流数据的聚类分析中.实验结果表明,EWSSC 算法对于高维数据流可以得到与批处理软子空间聚类方法近似一致的实验结果.     

k-means型软子空间聚类算法

软子空间聚类是聚类研究领域的一个重要分支和研究热点。高维空间聚类以数据分布稀疏和"维度效应"现象等问题而成为难点。在分析现有软子空间聚类算法不足的基础上,引入子空间差异的概念;在此基础上,结合簇内紧凑度的信息来设计新的目标优化函数;提出了一种新的k-means型软子空间聚类算法,该算法在聚类过程中无需设置额外的参数。理论分析与实验结果表明,相对于其他的软子空间算法,该算法具有更好的聚类精度。     

Subspace identification methods for Hammerstein systems: rank constraint and dimension problem

This article studies the subspace identification methods (SIMs) for Hammerstein systems with major focus on a rank constraint and the related dimension problem. We analyse the effects of the rank constraint on the three steps of a unifying framework for SIMs: the rank constraint has no effect on the first two steps, but does so on the third step. If the rank constraint is ignored, as in the existing over-parametrised method (OPM) for Hammerstein system identification, the optimality of the resulting estimate can still be established. Even so, the OPM may suffer from the dimension problem resulting in a low numerical efficiency. To resolve the dimension problem, we propose a new subspace-based method, named as the least-parametrised method (LPM), for identification of Hammerstein systems with non-coupling input nonlinearities. Simulation results are provided to demonstrate the effectiveness of the LPM, and show the necessity of considering the rank constraint to improve the numerical efficiency.     

基于扩展PASTd跟踪算法的谐波检测方法

子空间分解类算法在理论上具有任意的高分辨率,非常适合于电力系统各类谐波的分析,但需要对高维矩阵进行特征值分解,这不仅费时而且不易于工程实现。将投影近似子空间跟踪算法引入电力系统谐波分析领域,详细分析评估PASTd算法的性能。仿真结果表明,紧缩投影近似子空间跟踪算法即PASTd算法不仅保留了子空间分解类算法的超分辨率特性,而且收敛速度较快,稳定性好,可推广用于电力系统谐波检测领域。     

一种基于双子空间的人脸美感分析方法

子空间技术是一种有效的人脸美感本征描述方法.为了克服单一子空间在人脸图像美感描述方面的不足,提出了一种基于主成分分析(PCA)与广义矩阵低秩逼近(Generalized low rank approximation matrix,GLRAM)双子空间的自动人脸美感分析方法.通过组合PCA和GLRAM子空间获取人脸美感特性的全局及局部本征描述,并利用高斯场模型(Gaussian field model,GF)构造组合子空间的内在几何结构关系.实验选用了一个光照、背景、表情、年龄和种族等变化比较显著的数据库,结果表明,提出的基于双子空间算法优于基于单一子空间的人脸美感分析方法.     

应用遗传算法优化子空间的SVM分类算法

提出了一种应用遗传算法优化子空间的SVM分类算法GS-SVM。该算法首先改进样本选择策略,采用基于置信度和凸包的样本选择方法,考虑类间距离和样本分布等因素,选择典型代表样本作为SVM的新训练集;然后采用矩阵式混合编码方式,利用遗传算法一并优化代表样本的特征子空间和SVM分类参数,并根据特征优化后的代表样本,构建SVM分类模型。在UCI的11个数据集上进行的仿真实验结果表明,该算法在大部分数据集上均可获得较小的样本规模和特征维数,以及较高的分类精度。     

结构约束的对称低秩表示子空间聚类算法

通过子空间聚类可获得高维数据的潜在子空间结构,但现有算法不能同时揭示数据全局低秩结构和局部稀疏结构特性,致使聚类性能受限.提出一种结构约束的对称低秩表示算法用于子空间聚类.在目标函数中添加结构约束和对称约束来限制低秩表示解的结构,构造一个加权稀疏和对称低秩的亲和度图,在此基础上,结合谱聚类方法实现高效的子空间聚类.实验...     

基于神经网络的运动参数估计

分别就两种约束使用神经网络对三维刚体运动进行参数估计.一是基于三维点匹配,将预测的运动参数作用于运动前的坐标,与运动后坐标进行比较;二是基于二维运动场,将使用预测的运动参数计算得出的二维运动场与图像序列中计算得出的二维运动场进行比较.两个神经网络均使用Newton-Raphson方法更新权值,以达到目标误差最小化.通过实验验证了该神经网络方法.     

Design and implementation of fractional differentiators,Part I: System based methods

The online calculation of fractional derivative is an important technique used in the control engineering area. To achieve satisfactory computing performance, this paper proposes a useful and effective tool named fractional order tracking differentiator (FOTD), which is easy to be implemented. A constructive method for designing the FOTD is developed using a class of asymptotically stable fractional order systems. It is shown that FOTD is the pioneering one to achieve it with simple structure, convenient construction and predominant performance. Finally, several examples are conducted to illustrate the effectiveness and efficiency of the proposed methods.     

高效多子空间Skyline查询处理算法

随着Skyline查询应用的增多,子空间Skyline查询成为热点。针对实际应用中用户从多角度审视某一数据集的需求,充分研究了多子空间Skyline查询问题。在分析现有子空间Skyline查询算法解决该问题不足的基础上,提出了子空间立方体群(subspace skycube group,SSG)结构,并给出了基于该结构的同时计算任意多个子空间Skyline查询的MSSC(multiple subspace skycube)算法。该算法采用子空间候选集(subspace candidate sets,SCS),并充分利用了子空间立方体群结构中各子空间Skyline结果间的共享关系;在此基础上,算法采用求和过滤以及最大值过滤等方法,对数据集进行剪枝和过滤,从而进一步提高算法效率。最后,分别用人造数据和真实数据对算法进行实验,并与现有算法进行比较,结果表明MSSC算法可以高效地解决多子空间Skyline查询问题。     

TLWCC：一种双层子空间加权协同聚类算法

协同聚类是对数据矩阵的行和列两个方向同时进行聚类的一类算法。本文将双层加权的思想引入协同聚类,提出了一种双层子空间加权协同聚类算法(TLWCC)。TLWCC对聚类块(co-cluster)加一层权重,对行和列再加一层权重,并且算法在迭代过程中自动计算块、行和列这三组权重。TLWCC考虑不同的块、行和列与相应块、行和列中心的距离,距离越大,认为其噪声越强,就给予小权重;反之噪声越弱,给予大权重。通过给噪声信息小权重,TLWCC能有效地降低噪声信息带来的干扰,提高聚类效果。本文通过四组实验展示TLWCC算法识别噪声信息的能力、参数选取对算法聚类结果的影响程度,算法的聚类性能和时间性能。     

一种平稳子空间分析的快速不动点算法

平稳子空间分析是新近发展的一种信号处理和数据分析技术,能够从观测到的多维非平稳信号中分离出平稳源信号。标准的平稳子空间分析算法基于Stiefel流形上的梯度下降方法。针对该算法收敛慢、耗时多的缺陷,根据关于Stiefel流形上优化问题的一阶最优性条件构造了迭代公式,提出一种新的平稳子空间分析的不动点算法。仿真实验表明,本文算法能够有效地分离出平稳源信号,分离性能优于已有的平稳子空间分析的不动点算法;与标准的基于Stiefel流形上梯度下降的算法相比,本文算法收敛更快,耗时更少。