高光谱影像概率分类向量机分类方法研究

从分析基于支持向量机和相关向量机的高光谱影像分类方法的优势和不足出发,将基于概率分类向量机的方法用于高光谱影像分类试验。在贝叶斯理论框架下,概率分类向量机为基函数权值引入截断Gauss先验概率分布,使得不同类别的基函数权值具有不同符号的先验分布,并利用EM算法进行参数推断,得到足够稀疏的概率模型,弥补了相关向量机选取错误类别的样本作为相关向量的不足,从而有效地提高了模型的分类精度和稳定性。OMIS和PHI影像分类试验表明,概率分类向量机能够很好地应用在高光谱影像分类。     

并行处理技术在地理信息数据处理中的应用

随着地理信息存储量的飞速增长,传统的单进程、集中式的数据处理方式已不能满足基于网络的地理信息服务的效能要求。分析对比了OpenMP,MPI和MapReduce等主流并行编程模式,将关系型数据库与分布式空间数据管理系统相结合,提出了面向并行处理的地理信息存储模型和数据组织模型,将该模型与传统模型进行了对比分析,并基于MapReduce实现了地理空间数据并行处理框架,选取了矢量数据装载、影像数据装载以及数据切片作为典型数据处理案例开展对比实验,该技术方案的处理效率均数倍于传统技术方案。实验表明,该模型能够很好地支持并行处理框架,可为分布式环境下数据处理中心构建提供一个有效解决方案。     

面向WebGL的矢量数据三维绘制技术

矢量数据的叠加显示能够提高三维虚拟地球的表达效果与分析能力。受限于GPU的计算精度,在三维虚拟地球中矢量数据绘制普遍存在抖动现象和深度冲突现象。对基于WebGL的矢量数据三维绘制中计算精度问题进行了分析,提出了使用CPU RTC技术和GPU RTE技术提高顶点变换的精度,使用多视锥渲染算法和深度平面技术解决深度缓存精度问题。实验证明,这几种技术和算法可以有效缓解抖动现象和深度冲突现象导致的视觉干扰,改善了各种尺度和范围的矢量数据在三维地形上的叠加显示效果。     

局部区域模型重力异常快速算法

地球重力场位系数模型可以用于计算局部重力扰动场元。然而随着地球重力场模型阶次的提高、局域重力场计算范围的增大,其计算速度往往不能满足工程需求。针对这一问题,在对位系数模型泰勒级数展开的基础上提出了采用向量运算、混合编程的方法,同时对连带勒让德函数Belikov递推方法中与经纬度无关的量进行了预先计算,有效提高了计算速度。提出的方法对于利用超高阶次重力场模型快速解算大范围、高分辨率重力场元数据以及累加求和计算具有一定的参考与借鉴意义。     

Smoothing and re-roughening processes: The geometric evolution of a single fault zone

The geometry of a fault zone exerts a major control on earthquake rupture processes and source parameters. Observations previously compiled from multiple faults suggest that fault surface shape evolves with displacement, but the specific processes driving the evolution of fault geometry within a single fault zone are not well understood. Here, we characterize the deformation history and geometry of an extraordinarily well-exposed fault using maps of cross-sectional exposures constructed with the Structure from Motion photogrammetric method. The La Quinta Fault, located in southern California, experienced at least three phases of deformation. Multiple layers of ultracataclasite formed during the most recent phase. Crosscutting relations between the layers define the evolution of the structures and demonstrate that new layers formed successively during the deformation history. Wear processes such as grain plucking from one layer into a younger layer and truncation of asperities at layer edges indicate that the layers were slip zones and the contacts between them slip surfaces. Slip surfaces that were not reactivated or modified after they were abandoned exhibit self-affine geometry, preserving the fault roughness from different stages of faulting. Roughness varies little between surfaces, except the last slip zone to form in the fault, which is the smoothest. This layer contains a distinct mineral assemblage, indicating that the composition of the fault rock exerts a control on roughness. In contrast, the similar roughness of the older slip zones, which have comparable mineralogy but clearly crosscut one another, suggests that as the fault matured the roughness of the active slip surface stayed approximately constant. Wear processes affected these layers, so for roughness to stay constant the roughening and smoothing effects of fault slip must have been approximately balanced. These observations suggest fault surface evolution occurs by nucleation of new surfaces and wear by competing smoothing and re-roughening processes.     

Characteristics of the Asian–Pacific Oscillation in Boreal Summer Simulated by BCC_CSM with Different Horizontal Resolutions

The summer Asian–Pacific Oscillation(APO) is a major teleconnection pattern that reflects the zonal thermal contrast between East Asia and the North Pacific in the upper troposphere. The performance of Beijing Climate Center Climate System Models(BCC CSMs) with different horizontal resolutions, i.e., BCC CSM1.1 and BCC CSM1.1(m), in reproducing APO interannual variability, APO-related precipitation anomalies, and associated atmospheric circulation anomalies, is evaluated.The results show that BCC CSM1.1(m) can successfully capture the interannual variability of the summer APO index. It is also more capable in reproducing the APO's spatial pattern, compared to BCC CSM1.1, due to its higher horizontal resolution. Associated with a positive APO index, the northward-shifted and intensified South Asian high, strengthened extratropical westerly jet, and tropical easterly jet in the upper troposphere, as well as the southwesterly monsoonal flow over North Africa and the Indian Ocean in the lower troposphere, are realistically represented by BCC CSM1.1(m), leading to an improvement in reproducing the increased precipitation over tropical North Africa, South Asia, and East Asia, as well as the decreased precipitation over subtropical North Africa, Japan, and North America. In contrast, these features are less consistent with observations when simulated by BCC CSM1.1. Regression analysis further indicates that surface temperature anomalies over the North Pacific and the southern and western flanks of the Tibetan Plateau are reasonably reproduced by BCC CSM1.1(m), which contributes to the substantial improvement in the simulation of the characteristics of summer APO compared to that of BCC CSM1.1.     

多普勒雷达观测在北京7.21暴雨数值模拟中的三维变分同化研究

基于背景误差的特征长度理论,研究调整背景误差水平分辨率对多普勒雷达资料三维变分同化的影响。首先利用NMC方法针对暴雨落区统计不同水平分辨率的背景误差协方差,分析两种不同分辨率的背景误差的结构特征,研究水平分辨率对背景误差特征长度的影响。将其应用于雷达资料同化中,研究背景误差水平分辨率变化对雷达资料同化的影响。结果表明:背景误差水平分辨率由27 km提高到3 km时,在大气低层体现出更细致的动力场信息,其动力场水平特征长度按水平分辨率的二次根递减,而温度场与水汽场水平特征长度变化不明显。在将不同分辨率的背景误差用于三维变分同化时,更高分辨率的背景误差可以在分析场增量中体现更细致的中小尺度信息,能够明显改善雷达径向速度资料同化效果,并在随后的暴雨数值模拟中雨量及其分布形态更接近实况。     

纹理特征向量与最大化熵法相结合的SAR影像非监督变化检测

合成孔径雷达(SAR)影像具有明显的斑点噪声,在变化检测中,一般需要考虑空间邻域信息。本文结合SAR影像丰富的纹理信息,提出一种考虑空间邻域信息的高分辨率SAR影像非监督变化检测方法,用基于灰度共生矩阵(GLCM)的32维纹理特征向量构造差异影像。通过最大化熵法自动选取阈值,对精度指标随窗口大小的变化进行回归分析,得到适合于变化检测的窗口为11×11。试验表明,本文方法优于马尔科夫随机场法,可以减小斑点噪声的影响,有效提高高分辨率SAR影像变化检测的精度。     

基于 GAMA 平台的多智能体应急疏散仿真模型

应急疏散的研究对象是一个复杂的社会系统,利用多智能体模型可以有效模拟人群复杂的社会行为。将多智能体模型应用于应急疏散领域,设计5种类型的智能体实体,构建多智能体应急疏散仿真模型,从被疏散个体的相互作用行为入手,研究整个被疏散群体的行为特征。基于GAMA平台,实现以矢量地理空间数据为基础的多智能体应急疏散仿真模型,通过仿真实验,动态呈现整个应急疏散过程。对灾害、突发事件应急疏散指挥,应急疏散设施建设、城市规划设计等工作具有重要的意义。