结合稀疏编码和空间约束的红外图像聚类分割研究

提出了结合稀疏编码和空间约束的红外图像聚类分割新算法, 在稀疏编码的基础上融合聚类算法, 扩展了传统的基于K-means聚类的图像分割方法. 结合稀疏编码的聚类分割算法能有效融合图像的局部信息, 便于利用像素之间的内在相关性, 但是对于分割会出现过分割和像素难以归类的问题.为此, 在字典的学习过程中, 将原子的聚类算法引入其中, 有助于缩减字典中原子所属类别的数目, 防止出现过分割; 考虑到像素及其邻域像素具有类别属性一致性的特点, 引入了空间类别属性约束信息, 并给出了一种交替优化算法. 联合学习字典、稀疏系数、聚类中心和隶属度, 将稀疏编码系数同原子对聚类中心的隶属程度相结合, 构造像素归属度来判断像素所属的类别. 实验结果表明, 该方法能够有效提高红外图像重要区域的分割效果, 具有较好的鲁棒性. 关键词： 图像分割 稀疏编码 聚类 空间约束     

Convex relaxation based sparse algorithm for hyperspectral target detection

Target detection in hyperspectral images is an important task. In this paper, we propose a sparsity based algorithm for target detection in hyperspectral images. In sparsity model, each hyperspectral pixel is represented by a linear combination of a few samples from an overcomplete dictionary, and the weighted vector for such reconstruction is sparse. This model has been applied in hyperspectral target detection and solved with several greedy algorithms. As conventional greedy algorithms may be trapped into a local optimum, we consider an alternative way to regularize the model and find a more accurate solution to the model. The proposed method is based on convex relaxation technique. The original sparse representation problem is regularized with a properly designed weighted ?₁ minimization and effectively solved with existing solver. The experiments on synthetic and real hyperspectral data suggest that the proposed algorithm outperforms the classical sparsity-based detection algorithms, such as Simultaneous Orthogonal Matching Pursuit (SOMP) and Simultaneous Subspace Pursuit (SSP) and conventional ?₁ minimization.     

Mutated and converged damping factors in lens system optimization to find new local minima

In this paper, we proposed a strategy to find new local minima based on mutated damping factors and converged damping factor (conventional damping factor). In this proposed strategy, the converged damping factor is used to make the algorithm converge to a certain local minimum, and the mutated damping factors are used to make the trapped solution jump out of present convergence area. We study the behavior of the two kinds of factors in lens system optimization. The proposed strategy is successfully applied in lens system optimization. The result shows that the proposed strategy is reliable to make a further improvement especially when the solution is approaching target design.     

A Regularization Homotopy Strategy for the Constrained Parameter Inversion of Partial Differential Equations

The main difficulty posed by the parameter inversion of partial differential equations lies in the presence of numerous local minima in the cost function. Inversion fails to converge to the global minimum point unless the initial estimate is close to the exact solution. Constraints can improve the convergence of the method, but ordinary iterative methods will still become trapped in local minima if the initial guess is far away from the exact solution. In order to overcome this drawback fully, this paper designs a homotopy strategy that makes natural use of constraints. Furthermore, due to the ill-posedness of inverse problem, the standard Tikhonov regularization is incorporated. The efficiency of the method is illustrated by solving the coefficient inversion of the saturation equation in the two-phase porous media.     

Computing Freidlin’s Cycles for the Overdamped Langevin Dynamics. Application to the Lennard-Jones-38 Cluster

The large time behavior of a stochastic system with infinitesimally small noise can be described in terms of Freidlin’s cycles. We show that if the system is gradient and the potential satisfies certain non-restrictive conditions, the hierarchy of cycles has a structure of a full binary tree, and each cycle is exited via the lowest saddle adjacent to it. Exploiting this property, we propose an algorithm for computing the asymptotic zero-temperature path and building a hierarchy of Freidlin’s cycles associated with the transition process between two given local equilibria. This algorithm is suitable for systems with a complex potential energy landscape with numerous minima. We apply it to find the asymptotic zero-temperature path and Freidlin’s cycles involved into the transition process between the two lowest minima of the Lennard-Jones cluster of 38 atoms. D. Wales’s stochastic network of minima and transition states of this cluster is used as an input.     

基于改进离散粒子群算法的测试优化选择

为了解决复杂系统测试性设计过程中测试选择所产生的组合爆炸问题,提出一种改进离散粒子群算法的智能方法。首先,为保证初始种群的多样性,利用混沌不重复遍历的特性初始化种群的速度和位置;其次,根据启发式规则和罚函数的方法计算粒子适应度,使算法具有良好的搜索性能;最后,通过采用自适应调整策略的惯性权重,使粒子易于跳出局部最优解,找到最优解。通过仿真实例验证了本文方法的有效性,优化结果满足系统各项测试性指标要求,可为复杂系统的测试优化选择提供有效指导。     

Fault Feature Extraction for Reciprocating Compressors Based on Underdetermined Blind Source Separation

In practical engineering applications, the vibration signals collected by sensors often contain outliers, resulting in the separation accuracy of source signals from the observed signals being seriously affected. The mixing matrix estimation is crucial to the underdetermined blind source separation (UBSS), determining the accuracy level of the source signals recovery. Therefore, a two-stage clustering method is proposed by combining hierarchical clustering and K-means to improve the reliability of the estimated mixing matrix in this paper. The proposed method is used to solve the two major problems in the K-means algorithm: the random selection of initial cluster centers and the sensitivity of the algorithm to outliers. Firstly, the observed signals are clustered by hierarchical clustering to get the cluster centers. Secondly, the cosine distance is used to eliminate the outliers deviating from cluster centers. Then, the initial cluster centers are obtained by calculating the mean value of each remaining cluster. Finally, the mixing matrix is estimated with the improved K-means, and the sources are recovered using the least square method. Simulation and the reciprocating compressor fault experiments demonstrate the effectiveness of the proposed method.     

Corners detection on finger vein images using the improved Harris algorithm

An improved Harris corner detection algorithm is proposed based on Barron operator, since Harris corner detection algorithm has a poor accuracy in positioning complex corner detection and may miss certain real corners. Firstly, the image gradient is calculated by using Barron operator to reduce the calculation errors from Prewitt operator or Sobel operator. Secondly, the centre B-spline function is used to smooth image, filter noise, and retain the corners information better. Thirdly, a non-maximal inhibition and corners sieving method is used to determine whether the detected corners are real corners or not. A square window is centered at the pixel and eliminate the corner if the value of the corner response function is non-maximal in the window. And then divide the test image into several blocks so as to process each block independently, and use a cyclic iterative method to determine the threshold value to make sure that the real corners are accurately selected. Finally, experiments indicate the algorithm has relatively great noise proof ability and is able to extract complex corners effectively.     

A fully automated algorithm under modified FCM framework for improved brain MR image segmentation

Automated brain magnetic resonance image (MRI) segmentation is a complex problem especially if accompanied by quality depreciating factors such as intensity inhomogeneity and noise. This article presents a new algorithm for automated segmentation of both normal and diseased brain MRI. An entropy driven homomorphic filtering technique has been employed in this work to remove the bias field. The initial cluster centers are estimated using a proposed algorithm called histogram-based local peak merger using adaptive window. Subsequently, a modified fuzzy c-mean (MFCM) technique using the neighborhood pixel considerations is applied. Finally, a new technique called neighborhood-based membership ambiguity correction (NMAC) has been used for smoothing the boundaries between different tissue classes as well as to remove small pixel level noise, which appear as misclassified pixels even after the MFCM approach. NMAC leads to much sharper boundaries between tissues and, hence, has been found to be highly effective in prominently estimating the tissue and tumor areas in a brain MR scan. The algorithm has been validated against MFCM and FMRIB software library using MRI scans from BrainWeb. Superior results to those achieved with MFCM technique have been observed along with the collateral advantages of fully automatic segmentation, faster computation and faster convergence of the objective function.     

Acceleration of Global Optimization Algorithm by Detecting Local Extrema Based on Machine Learning

This paper features the study of global optimization problems and numerical methods of their solution. Such problems are computationally expensive since the objective function can be multi-extremal, nondifferentiable, and, as a rule, given in the form of a “black box”. This study used a deterministic algorithm for finding the global extremum. This algorithm is based neither on the concept of multistart, nor nature-inspired algorithms. The article provides computational rules of the one-dimensional algorithm and the nested optimization scheme which could be applied for solving multidimensional problems. Please note that the solution complexity of global optimization problems essentially depends on the presence of multiple local extrema. In this paper, we apply machine learning methods to identify regions of attraction of local minima. The use of local optimization algorithms in the selected regions can significantly accelerate the convergence of global search as it could reduce the number of search trials in the vicinity of local minima. The results of computational experiments carried out on several hundred global optimization problems of different dimensionalities presented in the paper confirm the effect of accelerated convergence (in terms of the number of search trials required to solve a problem with a given accuracy).     

分子构象聚类中的K-means密度峰值搜索算法

分子构象的聚类是搜索分子动力学模拟轨迹中代表构象的主要方法。 它是分析复杂构象改变或分子间相互作用机制的关键步骤. 作为一种基于密度的聚类算法,密度峰值搜索算法因其聚类的准确度而被应用于分子聚类过程中. 但随着模拟时长的增长,密度峰值搜索算法较低的计算效率限制了其应用的可能. 本文提出K-means密度峰值搜索算法的聚类算法,它是密度峰值搜索算法在计算效率方面的一个扩展版本,用于解决密度峰值搜索算法中巨大的资源消耗问题. 在K-means密度峰值搜索算法中,首先,通过高效的聚类算法(例如K-means)进行初始聚类,得到的聚类中心被定义为具有权重的典型点. 然后,对加权的典型点通过密度峰值搜索算法实现二次聚类,并细化点为核心点、边界点、加细光晕点. 在与密度峰值搜索算法具有相似的精度的同时,计算复杂度由O(n²)降至O(n). 通过二面角,二级结构,关联图描述的分子构象,将KFDP用于多个模拟轨迹的聚类过程中. 并通过与K-means聚类算法,DBSCAN聚类算法的比较结果,验证了K-means密度峰值搜索算法的优势.     

基于伪魏格纳分布分解的合成孔径雷达图像目标与阴影分割

在二维魏格纳分布的框架内,针对魏格纳变换的交叉项问题和计算量大的问题,提出了合成孔径雷达图像局部伪魏格纳变换的目标和目标阴影的分割方法.首先,将合成孔径雷达图像进行二维伪魏格纳变换,得到各像素点的二维能量谱图|然后提取各像素点的二维能量谱图对应位置值形成多个不同频段的与原图像同大小的能量谱图|最后,对不同频段的能量谱图采用不同的处理方法后,将各能量谱图相加处理后形成区域标识图像,最终得到原图像的目标和目标阴影分割图像.本文利用该方法对MSTAR切片图像进行了分割试验,并对分割图像与频谱最大值距离或方位分割算法和基于双参量CFAR与隐马尔科夫联合分割算法进行了分割图像对比度对比.实验结果表明,采用本文算法的合成孔径雷达分割图像,对比度明显提高,且保留了目标图像细节.     

Aging without disorder on long time scales

We study the Metropolis dynamics of a simple spin system without disorder, which exhibits glassy dynamics at low temperatures. We use an implementation of the algorithm of Bortz, Kalos and Lebowitz [1]. This method turns out to be very efficient for the study of glassy systems, which get trapped in local minima on many different time scales. We find strong evidence of aging effects at low temperatures. We relate these effects to the distribution function of the trapping times of single configurations.     

Chaotic Dynamics in Weight Space of Neural Networks

When a special nonlinear self-feedback term is introduced into the dynamical equation of the backpropagation training algorithm for networks, the dynamics in weight space of networks can become chaotic. Chaotic dynamics of the system can help it escape from the most commonplace local minima of the energy. Simulation on the XOR problem and the prediction of chaotic time series have shown that the proposed chaotic training algorithm can converge to the global minimum or its approximate solutions efficiently and dramatically faster than the original backpropagation training algorithm.     

Domains in multiband superconductors

Multiband superconductors can have several types of domains that are inhibited in conventional single-band superconductors. These domains are phase domains and chiral domains and their domain wall are an interband phase difference soliton. In a superconductor with an odd number of electronic bands (five or more) and with positive interband Josephson interactions, we find other types of domains with different interband phase differences. We call these domains configuration domains because pseudo-order parameters for each band are dispersed in the complex plain and several configurations, which have several local minima. Fractional vortices serve as hubs for phase difference solitons (configuration domain walls). The divergence of the number of configurations with local minima would pose a serious problem for the stability of superconductivity.     

Two sub-pixel processing algorithms for high accuracy particle centre estimation in low seeding density particle image velocimetry

This article presents two algorithms for spatial processing of low seeding density PIV (particle image velocimetry) images which lead to sub-pixel precision in particle positioning. The particle centres are estimated to accuracies of the order of 0.1 pixel, yielding 1% error in velocity calculation. The first algorithm discriminates valid particles from the rest of the image and determines their centres in Cartesian coordinates by using a two-dimensional Gaussian fit. The second algorithm performs local correlation between particle pairs and determines instantaneous two-dimensional velocities. The methods have been applied initially to simulated data. Gaussian noise and distortion has then been added to simulate experimental conditions. It is shown that, in comparison with conventional methods, the new algorithms offer up to an order of magnitude higher accuracy for particle centre estimation. Finally, the Gaussian fit approach has been used to map an experimental transonic flow field from the stator trailing edge wake region of a cascade with an estimated error of 1%. The experimental results are found to be in good agreement with previous theoretical steady-state viscous calculations.     

基于局部线性变换的色阶映射算法

为了提高高动态图像显示的视觉效果,提出了一种基于局部线性变换的色阶映射算法。针对图像直接线性压缩能力较差的缺点,本文提出一种局部的线性变换模型。由于局部窗的结构,是利用局部窗的中心像素点与窗内的其他像素点的均值,方差的关系构造而成,因此基于局部线性变换的色阶映射算法能够抑制由具有高对比度的图像边缘所引起的鬼影和光晕现象。首先,根据线性变换将图像的色阶映射转换为求解泛函的最优解;其次,通过推导求解出线性变换的参数并将泛函最优化问题转换为求解一个线性方程组问题;最后求解线性方程组,得出视觉效果较好的低动态范围图像。有效避免传统色阶映射出现的光晕和亮度不连续的现象,能够更多的保留高动态图像的细节。最后通过实验结果对比分析,通过信息熵和对比度两个参数可以体现文中算法较好。     

Algorithm of far-field centre estimation based on phase-only matched filter

Estimation of the far-field centre is carried out in beam auto-alignment. In this paper, the features of the far-field of a square beam are presented. Based on these features, a phase-only matched filter is designed, and the algorithm of centre estimation is developed. Using the simulated images with different kinds of noise and the 40 test images that are taken in sequence, the accuracy of this algorithm is estimated. Results show that the error is no more than one pixel for simulated noise images with a 99% probability, and the stability is restricted within one pixel for test images. Using the improved algorithm, the consumed time is reduced to 0.049s.     

Quantum dynamics of phonon assisted emission of carriers at deep level centers

A formalism is proposed to investigate quantum dynamics of localized states involving highly non-adiabatic time-evolution of electron-lattice systems. The effect of electron itinerancy is projected onto the dynamics of local variables through an integral kernel of Volterra's integral equation. The method is applied to the problem of thermal emission of carriers at deep level centers in semiconductors. It is shown that the real situation is in the adiabatic limit, and the probability of thermal emission of the trapped carriers is one per a single lattice oscillation, if the amplitude of the oscillation exceeds a critical value but zero if not.