基于不变性矩特征的人工神经网络2-D形状识别

在2—D不变性矩快速计算的基础上，利用人工神经网络，使用不变性矩完成2—D形状的识别。实验结果与现有文献的结果比较，本文采用的方法具有特征数目少、意义统一、识别快速、准确的特点。     

On the Sitnikov problem

A mapping which reflects the properties of the Sitnikov problem is derived. We study the mapping instead of the original differential equations and discover that there exists a hyperbolic invariant set. The theoretical prediction of the disorder region agrees remarkably with numerical results. We also discuss the LCEs and KS-entropy of the dynamical system.This project is supported by the National Science Foundation of China.     

Stability Domain and Invariant Manifolds of 2d Area-Preserving Diffeomorphisms

We study the stability domain of generic 2D area-preserving polynomialdiffeomorphisms. The starting point of our analysis is the study of thedistribution of stable and unstable fixed points. We show that the locationof fixed points and their stability type are linked to the degree of thepolynomial map. These results are based on a classification Theorem forplane automorphisms by Friedland and Milnor. Then we discuss the problem ofdetermining the domain in phase space where stable motion occurs. We showthat the boundary of the stability domain is given by the invariantmanifolds emanating from the outermost unstable fixed point of low period(one or two). This fact extends previous results obtained for reversiblearea-preserving polynomial maps of the plane. This analysis is based onanalytical arguments and is supported by the results of numericalsimulations.     

Motion close to the Hopf Bifurcation Of the vertical family of periodic orbits of L4

The paper deals with different kinds of invariant motions (periodic orbits, 2D and 3D invariant tori and invariant manifolds of periodic orbits) in order to analyze the Hamiltonian direct Hopf bifurcation that takes place close to the Lyapunov vertical family of periodic orbits of the triangular equilibrium point L₄ in the 3D restricted three-body problem (RTBP) for the mass parameter, μ greater than (and close to) μ_R (Routh’s mass parameter). Consequences of such bifurcation, concerning the confinement of the motion close to the hyperbolic orbits and the 3D nearby tori are also described.     

Wisdom system: Dynamics in the adiabatic approximation

A simple approximate model of the asteroid dynamics near the 3:1 mean–motion resonance with Jupiter can be described by a Hamiltonian system with two degrees of freedom. The phase variables of this system evolve at different rates and can be subdivided into the ‘fast’ and ‘slow’ ones. Using the averaging technique, wisdom obtained the evolutionary equations which allow to study the long-term behavior of the slow variables. The dynamic system described by the averaged equations will be called the ‘Wisdom system’ below. The investigation of the, wisdom system properties allows us to present detailed classification of the slow variables’ evolution paths. The validity of the averaged equations is closely connected with the conservation of the approximate integral (adiabatic invariant) possessed by the original system. Qualitative changes in the behavior of the fast variables cause the violations of the adiabatic invariance. As a result the adiabatic chaos phenomenon takes place. Our analysis reveals numerous stable periodic trajectories in the region of the adiabatic chaos.     

KS－熵与混沌层次

有两种不同类型的混沌层次：一种得之于映射维数的增加；另一种则由不变环面分叉产生。通过计算相应的ＫＳ－熵，我们发现前者只是一种几何现象，后者则是本质的混沌层次，即伴随有混沌性的增加。     

Probability of capture for the second fundamental model of resonance

The extension of the adiabatic invariant theory for one-degree of freedom Hamiltonian systems with varying parameters proposes a way of evaluating the probability of capture from one domain of the phase space into another. We derive here analytic expressions for these probabilities for a typical model of resonance of orderN. We show that the probabilities depend only upon two parameters and not four as expected a priori.     

基于三维正演的音频大地电磁阻抗相位不变量校正技术

复杂地形、地质条件的大地电磁数据解释容易出现假象,采用三维正演技术模拟地形和地表不均匀体的背景响应,对实测数据阻抗相位不变量进行校正,实现更准确的定性分析;对三维异常体模型的合成数据进行一维、二维多参数反演试算,以确定地形剧变区选择反演技术的最佳方案.合成数据的试反演结果显示一维反演水平切片假异常较多,二维反演能压制测向假异常,但不能压制走向的假异常,水平切片多出现测向条带.使用本文提出的阻抗相位不变量校正法扣除地形、地表背景响应,结合一维、二维反演,能使实际资料解释成果更加可靠.     

小波域遥感图像不变纹理分类

分析了有限脊小波变换可以实现图像的旋转不变性和平移不变性,提出了结合两种小波变换提取图像纹理特征的方法,实现了在小波域中进行图像的不变纹理分类.     

结合特征分量构建和面向对象方法提取高分辨率卫星影像阴影

针对高分辨率卫星影像,提出一种特征分量构建与面向对象结合的阴影提取方法。分析遥感阴影光谱特性,构建彩色不变特征C3、亮度特征I、主成分第一特征量PC1以及蓝色波段和近红外波段归一化比率特征RATIOb_nir,增强阴影信息。采用线性变换将几个特征分量Digital Number(DN)值归一化到相同范围,对这几个分量进行综合分析。以I和PC1分量为输入对影像进行多尺度分割,建立包括波段均值、标准差、最大差异等特征的规则集,实现面向对象的阴影信息提取。选取20幅QuickBird影像为例进行阴影提取实验,平均总体精度为97%,平均用户精度为96%,平均Kappa系数为0.94。实验结果表明,相对传统基于像素信息提取方法,本文方法提取阴影斑块完整,无破碎图斑;相对基于原始光谱的面向对象方法,本文方法提取精度更高。