基于核各向异性扩散的红外小目标检测

为了减少红外图像中背景边缘对检测的影响,提出了一种具有鲁棒性的弱小目标检测算法,该算法利用核各向异性扩散模型进行背景预测,再与原图像差分实现弱小目标检测。为了提高算法的自适应能力,提出了一种鲁棒性扩散系数,能够根据图像背景的起伏程度自适应调整扩散系数曲线的陡峭程度。实验结果表明,与现有的检测算法相比,该算法能够在不同类型的复杂背景下有效抑制背景及其边缘,保留目标大小,降低虚警率,具有更强的鲁棒性。     

Infinitely many singular interactions on noncompact manifolds

We show that the ground state energy is bounded from below when there are infinitely many attractive delta function potentials placed in arbitrary locations, while all being separated at least by a minimum distance, on two dimensional non-compact manifold. To facilitate the reading of the paper, we first present the arguments in the setting of Cartan–Hadamard manifolds and then subsequently discuss the general case. For this purpose, we employ the heat kernel techniques as well as some comparison theorems of Riemannian geometry, thus generalizing the arguments in the flat case following the approach presented in Albeverio et al. (2004).     

基于优化核极限学习机的风电功率时间序列预测

针对时间序列预测,在单隐层前馈神经网络的基础上,基于进化计算的优化策略,提出了一种优化的核极限学习机(optimized kernel extreme learning machine,O-KELM)方法.与极限学习机(extreme learning machine,ELM)方法相比,核极限学习机(kernel extreme learning machine,KELM)方法无须设定网络隐含层节点的数目,以核函数表示未知的隐含层非线性特征映射,通过正则化最小二乘算法计算网络的输出权值,它能以极快的学习速度获得良好的推广性.在KELM的基础上,分别将遗传算法、模拟退火、微分演化三种进化算法用于模型的结构输入选择、正则化系数以及核参数的优化选取,以进一步提高网络的性能.将O-KELM方法应用于标准Mackey-Glass混沌时间序列预测及某地区的风电功率时间序列预测实例中,在同等条件下,还与优化的极限学习机(optimized extreme learning machine,O-ELM)方法进行比较.实验结果表明,所提出的O-KELM方法在预测精度上优于O-ELM方法,表明了其有效性.     

用于混沌时间序列预测的组合核函数最小二乘支持向量机

针对混沌时间序列的预测问题,考虑到单一核函数的最小二乘支持向量机无法明显提高预测精度,提出了一种组合核函数的最小二乘支持向量机预测模型,模型中采用多项式函数与径向基函数组合构建核函数.同时,还对遗传算法进行了改进,使之具有更快的收敛速度和更高的精度,改进的遗传算法适用于解决预测模型中的参数优化问题.通过典型的Lorenz时间序列、Mackey-Glass时间序列、太阳黑子数时间序列以及具有混沌特性的网络流量时间序列对该模型进行了验证.仿真结果表明所提出的模型是有效的.     

Exact solution of a coagulation equation with a product kernel in the multicomponent case

In this paper, we obtain the general solution for the continuous Smoluchowski equation in the multicomponent case with a product kernel as a series expansion. The solution of the problem involves the Laplace transform in several dimensions. We obtain a nonlinear partial differential equation (PDE) of the advective kind generalizing the one previously given by other authors for the mono-component case.As in its relative mono-component case, gelation is produced at some point, the conditions for its occurrence being the same as those for the mono-component case, though substituting a sum of derivatives by a derivative in the Laplace transform field. We demonstrate that for a multicomponent particle size distribution (PSD) of multiplicative form, it is sufficient for one of the marginal PSDs to generate instantaneous gelation for the occurrence of instantaneous gelation in the multicomponent PSD.The general solution is applied to several specific cases, a discrete case that recovers a previously known solution, and another two continuous cases which can be used to check numerical methods designed to directly solve the Smoluchowski equation in more general cases.We have compared the solutions for the multicomponent PSD for constant, additive and product kernels and we conjecture about the relation existing between the functional forms for the solutions both in the mono-component and the multicomponent case.Finally, we have analysed the shape of the solutions for multicomponent PSD for constant, additive and product kernels for very small masses of components, obtaining a qualitatively different behaviour for the product kernel. This has effects in the mixing state of the sol phase as time passes.     

Intelligibility enhancement for noisy whispered speech using asymmetric cost function

We proposed two whispered speech enhancement methods based on asymmetric cost functions in this paper to deal with the amplification and attenuation distortions of whispered speech distinctively.The modified Itakura-Saito（MIS）distance function provides more penalties to speech amplification distortion,whereas the Kullback-Leibler（KL）divergence function gives more penalties to speech attenuation distortion.The experimental results show that the MIS function based method achieves significant improvement of intelligibility in contrast to the conventional speech enhancement algorithms when the signal-to-noise ratio（SNR）falls below-6 dB,whereas the KL function based one achieves the similar result as the minimum mean square error（MMSE）speech enhancement method.The results show that the effects of the amplification and attenuation distortions on the intelligibility of the enhanced whisper are different,where larger attenuation distortion may result in better intelligibility of speech with low SNR.However,the attenuation distortion has small effects on intelligibility of speech with high SNR.     

On the spurious eigenvalues for a concentric sphere in BIEM

In this paper, the null-field integral equation method is employed to study the occurring mechanism of spurious eigenvalues for a concentric sphere. By expanding the fundamental solution into degenerate kernels and expressing the boundary density in terms of spherical harmonics, all boundary integrals can be analytically determined. It is noted that our null-field integral formulation can locate the collocation point on the real boundary thanks to the degenerate kernel. In addition, the spurious eigenvalues are parasitized in the formulations while true eigensolutions are dependent on the boundary condition such as the Dirichlet or Neumann problem. By using the updating term and updating document of singular value decomposition (SVD) technique, true and spurious eigenvalues can be extracted out, respectively. Besides, true and spurious boundary eigenvectors are obtained in the right and left unitary vectors in the SVD structure of the influence matrices. This finding agrees with that of 2D cases.     

正交饱和效应模型及其统计分析

文\cite{14}讨论了2水平正交饱和设计的一种统计分析方法---零效应搜索法, 本文将此方法推广到一般的饱和模型---正交饱和效应模型. 首先给出正交饱和效应模型的定义及其相关的参数估计和平方和分解, 然后用零效应搜索法对显著效应的检验问题进行分析, 模拟结果显示效果很好.     

基于Bayes估计的有限元模型修正

讨论了基于Bayes方法来修正有限元动力模型问题 ,给出了利用Bayes方法估计参数的公式 .根据Bayes估计结果 ,应用矩阵逆特征值理论 ,提出了有限元模型修正方法 .最后用一个数值试验说明了该方法的有效性 .     

一类弱奇异边值问题的大范围收敛算法

该文研究如下的弱奇异边值问题: (p(x)y')'=f(x, y),0b₀g(x), 0≤b₀<1, 边值条件为y(0)=A,αy(1)+β y'(1)=γ 或y'(0)=0,αy(1)+βy'(1)=γ (R.K.Pandey 和 Arvind K.Singh 给出了一种求解此问题的二阶有限差分方法^[1]. 在再生核空间中讨论方程解的存在性, 给出一种新的迭代算法,这种迭代算法是大范围收敛的. 给出数值算例并与R. K. Pandey 和Arvind K.Singh 给出的方法进行比较说明该文方法的有效性.