基于相关向量机的高光谱影像分类研究

虽然支持向量机在高光谱影像分类得到成功应用,但是它自身固有许多不足之处。相关向量机是在贝叶斯框架下提出的更加稀疏的学习机器,它没有规则化系数,其核函数不需要满足Mercer条件,不仅具备良好的泛化能力,而且还能够得到具有统计意义的预测结果。本文从分析支持向量机用于高光谱影像分类存在的不足出发,提出了一种基于相关向量机的高光谱影像分类方法,介绍了稀疏贝叶斯分类模型,将相关向量机学习转化为最大化边缘似然函数估计问题,并采用了快速序列稀疏贝叶斯学习算法。通过PHI和OMIS影像分类实验分析表明了基于相关向量机的高光谱影像分类方法的优越性。     

融合数据内部变化信息的丰度估计算法

丰度估计(AE)是从高光谱图像中识别地物的关键预处理技术。由于线性模型的可解释性以及数学上的可操作性,基于该模型的线性回归技术CLR(Constrained Linear Regression)在丰度估计中受到了广泛关注。目前,该方法仅仅考虑到了估计数据与被估计数据之间的能量相似性,没有考虑数据内部的变化信息之间的相似性,比如一阶梯度之间的相似性以及二阶梯度之间的相似性。为了提高丰度估计精度,本文提出了融合数据内部变化信息的稀疏低秩丰度估计算法。首先通过增加一阶梯度和二阶梯度的约束项改进传统的丰度估计的数学模型。其次,通过采用范数不等式和优化理论证明了在约束条件下,该模型的有效性及该模型在相关领域的可拓展性。接着,采用辅助变量将改进的数学模型变为增强拉格朗日函数。最后,采用交替双向乘子技术ADMM(Alternating Direction Method of Multipliers)求解该模型并估计高光谱图像的丰度。经仿真实验和实际高光谱图像的实验证明该方法能够改善仿真数据和实际高光谱数据的丰度估计的效果,特别是当端元的丰度存在丰富的变化细节时,丰度估计的精度和抗噪性能均优于当前较流行的丰度估计算法。     

Effect of unequal injection rates and different hopping rates on asymmetric exclusion processes with junction

In this paper, the effects of unequal injection rates and different hopping rates on the asymmetric simple exclusion process (ASEP) with a 2-input 1-output junction are studied by using a simple mean-field approach and extensive computer simulations. The steady-state particle currents, the density profiles, and the phase diagrams are obtained. It is shown that with unequal injection rates and different hopping rates, the phase diagram structure is qualitatively changed. The theoretical calculations are in good agreement with Monte Carlo simulations.     

Entanglement of two atoms interacting with a dissipative coherent cavity field without rotating wave approximation

Considering two identical two-level atoms interacting with a single-model dissipative coherent cavity field without rotating wave approximation, we explore the entanglement dynamics of the two atoms prepared in different states using concurrence. Interestingly, our results show that the entanglement between the two atoms that initially disentangled will come up to a large constant rapidly, and then keeps steady in the following time or always has its maximum when prepared in some special Bell states. The model considered in this study is a good candidate for quantum information processing especially for quantum computation as steady high-degree atomic entanglement resource obtained in dissipative cavity.     

优化点扩散函数估计与稀疏约束的图像盲复原

在卫星光学图像盲复原问题中,点扩散函数的估计和复原模型选择对结果影响很大。针对点扩散函数模板估计困难的问题,该文在分析图像中近似点状地物成像特点的基础上,提出了基于点状地物信息的椭圆抛物面模型点扩散函数估计方法,得到较好的点扩散函数模板;针对图像噪声对复原的影响问题,采用能够分离图像的信号和噪声的归一化稀疏约束模型进行盲复原。利用"吉林一号"卫星图像进行了实验。结果表明:该文提出的优化点扩散函数估计方法可行,盲复原后图像质量得到了提高。     

Seismic signal enhancement based on the low-rank methods

Based on the fact that the Hankel matrix constructed by noise-free seismic data is low-rank, low-rank approximation (or rank-reduction) methods have been widely used for removing noise from seismic data. Due to the linear-event assumption of the traditional low-rank approximation method, it is difficult to define a rank that optimally separates the data subspace into signal and noise subspaces. For preserving the most useful signal energy, a relatively large rank threshold is often chosen, which inevitably leaves residual noise. To reduce the energy of residual noise, we propose an optimally damped rank-reduction method. The optimal damping is applied via two steps. In the first step, a set of optimal damping weights is derived. In the second step, we derive an optimal singular value damping operator. We review several traditional low-rank methods and compare their performance with the new one. We also compare these low-rank methods with two sparsity-promoting transform methods. Examples demonstrate that the proposed optimally damped rank-reduction method could get significantly cleaner denoised images compared with the state-of-the-art methods.     

宽频数据在致密砂岩储层预测中的应用

西湖凹陷深层致密砂岩储层具有良好的勘探开发前景,受埋深影响,目的层地震资料品质较差.A构造通过斜缆宽频采集和处理获取宽频地震数据,提升了资料品质,然而应用常规子波提取方法对宽频数据进行子波提取并反演计算纵波阻抗,结果与井上实测数值差异较大,影响储层的定量解释.针对这一问题,提出统计性子波和确定性子波相结合的长短子波合并宽频子波提取方法,提取的宽频子波比常规子波低频丰富、旁瓣小,能更真实地反映地震信息,约束稀疏脉冲反演的纵波阻抗结果与测井曲线吻合度更高.基于宽频数据和常规数据分别进行约束稀疏脉冲弹性波阻抗反演,预测A构造优质储层分布,经已钻井证实,宽频数据比常规数据储层预测精度高,预测的储层展布特征与研究区地质沉积认识一致.结果表明:这种基于宽频子波提取的宽频资料应用方法有效降低了致密砂岩储层预测的多解性.     

A numerical study on deblending of land simultaneous shooting acquisition data via rank-reduction filtering and signal enhancement applications

We propose a workflow of deblending methodology comprised of rank-reduction filtering followed by a signal enhancing process. This methodology can be used to preserve coherent subsurface reflections and at the same time to remove incoherent and interference noise. In pseudo-deblended data, the blending noise exhibits coherent events, whereas in any other data domain (i.e. common receiver, common midpoint and common offset), it appears incoherent and is regarded as an outlier. In order to perform signal deblending, a robust implementation of rank-reduction filtering is employed to eliminate the blending noise and is referred to as a joint sparse and low-rank approximation. Deblending via rank-reduction filtering gives a reasonable result with a sufficient signal-to-noise ratio. However, for land data acquired using unconstrained simultaneous shooting, rank-reduction–based deblending applications alone do not completely attenuate the interference noise. A considerable amount of signal leakage is observed in the residual component, which can affect further data processing and analyses. In this study, we propose a deblending workflow via a rank-reduction filter followed by post-processing steps comprising a nonlinear masking filter and a local orthogonalization weight application. Although each application shows a few footprints of leaked signal energy, the proposed combined workflow restores the signal energy from the residual component achieving significantly signal-to-noise ratio enhancement. These hierarchical schemes are applied on land simultaneous shooting acquisition data sets and produced cleaner and reliable deblended data ready for further data processing.     

计算岩石波速空间平均的极限近似模型

本文在传统的岩石波速空间平均模型（Voight模型,Ruess模型,Hill模型和几何平均模型）的基础上,提出了一种极限近似模型.在这种模型中,它是利用作者提出的一个基于Hill模型（即代数平均模型）和几何平均模型的递推关系式,并利用这个关系式计算求出模型的极限近似值,该值介于Hill模型和几何平均模型之间,是具有典型代表意义的一个值.     

海洋磁力异常逼近方法研究

通过对常用的数值逼近方法的分析和研究,针对海洋磁力测量的特点,仿真计算分析了移动曲面法、Hardy多面函数法、Shepard法和Kriging法在不同情况下的插值精度。同时,给出了一个实例来计算分析四种逼近方法插值精度。仿真与实例计算结果表明,已知点的分布情况及磁异常变化情况不同时,四种逼近方法的插值精度是不同的。针对不同的情况,本文总结出了适合于海洋磁力异常逼近的方法。