多源空间数据融合方法的研究

数据融合是目前遥感应用领域的一个热点问题，本文介绍了传统的数据融合方法，以及最近比较常用的融合手段如基于附加小波分解的方法，并且提出将数字高程模型数据、地理信息系统数据、作为空间数据来源来进行基于神经网络的方法数据融合的设想。     

大地电磁阻抗张量正则分解的地质含义分析

由于常规的张量旋转分析方法的局限性，在处理三维阻抗问题时不能得到令人满意的结果.最新发展的三维阻抗张量的正则分解方法由于不需约束条件，可以自然的从三维阻抗张量中提取完整的参数体系，因而具有很好的应用前景.本文对正则分解的方法从理论上进行了分析和证明，对原有参数计算公式做适当的修改，将正则分解与坐标旋转结合起来，分析二者之间的联系与区别，并探讨参数的地质含义.     

N，N’-双（3-氨丙基）草酰胺合铜非等温热分解动力学研究

采用 TG- DTG- DTA热分析技术研究了 N,N’-双 (3-氨丙基 )草酰胺合胴 ( )配合物在动态空气气氛中的热行为 ;用微分法 (Achar法 )和积分法 (Coats- Redfen法 )协同处理非等温 TG数据 ,通过对比热分解动力学参数 E和 ln A,提出了配合物第一阶段热分解动力学方程 :dα/ dt =Ae- E/ RT (1-α ) 2 及第二阶段热分解动力学方程 :dα/ dt=Ae- E/ RT(1-α) ;并由动力学补偿效应获得了 E和 ln A的数学表达式     

分解算法及其收敛性质

本文给出了求解线性等式及不等式约束的非线性规划的一个新算法，并运用矩阵分解，给出了求逆的递推公式。该算法计算简单，证明了算法的收敛性及具有超线性收敛性质。     

小波变换及其在影像表示中的应用和其方向选择性算法的改进

小波理论是研究多分辨率信号处理的理论,目前被公认为有前途的新理论。金字塔结构的影像处理、影像理解,现代GIS处理多比例尺的要求以及地形、场论中的逼近均需有关这一理论的支持。本文利用小波变换分解原始影像以实现影像的小波表示,并发展为任意方向的小波变换、逆变换,利用它检测航片中建筑物及其阴影的公共边缘。这种表示方式可为某些影像特征或识别目标提供较为经济的分辨率,该表示方式使影像中具有一定层次的结构更易于发现和利用。实验结果证明了此方法是有效的。     

地震前兆信息系统的信息分解与浓缩

本文运用数值分类学中的因子分析方法,试图从多种前兆观测时序集中浓缩与分解出与地震孕育发生有关的成因信息。并用诸如“地震过程”这样一些概念来代表和归纳一组互相紧密相关的前兆现象和特征。这是与孕震系统在统一区域应力场作用下的整体性行为及子系统之间的协同关联运动的事实在思路上是相一致的。若干震例的研究结果表明,数值分类学是我们实现多台站、多手段,时、空动态综合预报思路的一种较好的方法。     

二维地层结构和速度的走时反演

介绍了在倾斜均匀各向同性层状介质情况下,根据地震反射走时反演二维地层结构和地层速度的地震射线层析成像方法——走时反演。正演模型射线追踪是根据Fermat原理实现的,即求解满足该原理的非线性方程组得到射线与界面的交点,进而计算相应的走时。反演是先假设一初始模型,用最优化方法使射线追踪走时与观察走时的残差极小。最后计算了分辨矩阵和信息密度矩阵,以评价反演结果。对有噪情形也进行了反演。     

Eigendecomposition of TDR waveforms: a novel method to determine water content and pore fluid concentration of sandy soils

In this study, an incident pulse signal of several harmonics (i.e., multiples of the fundamental frequency) was used as a source in the time domain reflectometry (TDR) probing technique. Reflected signals were captured by an oscilloscope and their characteristics were determined via eigendecomposition. Autoregressive modeling and singular value decomposition were used to calculate the eigenvalues and the most significant ones were identified based on power spectrum. A multivariate statistical analysis was performed for the two most dominant eigenvalues, which are dependent on water content and salt concentrations, and regression equations were obtained. To determine the water content and salt concentrations in terms of the first and second eigenvalues, a modified Powell hybrid algorithm was used to solve the obtained system of nonlinear equations. Actual and predicted results are in agreement indicating that the developed method is very successful in predicting water content and salt concentrations. Furthermore, on comparing the eigendecomposition method with Fourier spectral analysis, one can observe that the former is superior in predicting water content and salt concentrations.     

A RAM-economizing domain decomposition technique for regional high-resolution groundwater simulation

A major limitation of the use of high-resolution groundwater models on a regional scale for resource management by water companies is the excessive RAM requirements of such models which surpass the capacity of today's PCs. A strategy is presented to overcome this problem by using overlapping domain decomposition techniques. Furthermore, because very long computing time is the bottleneck for the practical use of this technique for large groundwater models, an analysis is also presented of a number of methods implemented to increase calculation performance. The approach presented here is characterized by a fairly simple structure that represents a generalized relaxation algorithm. It can be adapted for use with finite element as well as with finite difference methods. Electronic Publication     

Another form for LAMBDA method

The LAMBDA method that was proposed by Teunissen is in-troduced. Then, on the basis of both the back-sequential conditional LS technique and the upper-triangular Cholesky decomposition, another form for LAMBDA method is proposed.This new form for LAMBDA method has the same principle and calculation speed as the traditional LAMBDA method.