基于MST聚类的空间数据离群挖掘算法

空间离群是指空间邻域中属性特征值明显不同于其他对象的空间对象,空间数据离群挖掘能为人们提供很多有趣的信息,但空间数据具有复杂的拓扑关系、方位关系和度量关系等空间特征,传统的面向事务型数据库的离群挖掘算法并不适用于空间数据库。本文提出了基于MST(Minimum Spanning Tree,最小生成树)聚类的空间数据离群挖掘算法(SOM);有机结合了最小生成树理论与密度的方法,既体现了空间离群的局部特性,又体现了空间离群的孤立程度。该算法通过MST维护空间数据的基本空间结构特征,通过打断MST中最不一致的边形成MST聚类,不仅具有密度的聚类方法能够聚集非球状簇和分布不均的数据集的特点,而且聚类结果不依赖于用户参数的选择,因此,离群挖掘结果更合理。最后,通过实例数据,验证了该算法的有效性,它适用于大规模空间数据集的离群挖掘。     

Structure and dynamics of the stratosphere and mesosphere revealed by VHF radar investigations

Powerful VHF radars are capable of almost continuously monitoring the threedimensional velocity vector and the distribution of turbulence in the middle atmosphere, i.e. the stratosphere and mesosphere. Methods of radar investigations of the middle atmosphere are outlined and the basic parameters, mean and fluctuating velocities as well as reflectivity and persistency of atmospheric structures, are defined. Results of radar investigations are described which show that the tropopause level as well as a criterion on the stability of the lower stratosphere can be deduced. Besides mean wind velocities, VHF radars can measure instantaneous velocities due to acoustic gravity waves. The interaction of gravity waves with the background wind is discussed, and it is shown that cumulus convection is an effective source of gravity waves in the lower stratosphere. The vertical microstructure of the stratosphere, manifesting itself in thin stratified sheets in which temperature steps occur, is investigated by applying knowledge from investigations of the oceanic thermocline. Possible origins, like shear generation and lateral convection of the microstructure of the stratosphere, are discussed. Observations of gravity waves in the mesosphere are reviewed and their connection with turbulence structures is pointed out. Finally, some open questions which could be answered by further VHF radar investigations are summarized.     

采用最优生成树的正射影像镶嵌线快速智能检测

针对高分辨率遥感影像提出了一种基于最优生成树的镶嵌线快速智能检测方法。该方法利用影像亮度差异和梯度构建重叠区域差分影像并视其为带权无向图,以Bottleneck模型为基础,采用最优生成树生成方法智能检测最佳镶嵌线。该方法回避了迭代搜索过程,具有速度快、效率高的优点。试验结果表明,所提出的方法能够在保证镶嵌线质量的同时具有较高的效率,可有效解决正射影像镶嵌过程中镶嵌线的自动选择问题。     

Wind fluctuations near a cold vortex-tropopause funnel system observed by the MU radar

Vertical and temporal variations of three-dimensional wind velocity associated with an upper-tropospheric cold vortex-tropopause funnel system were observed by an MST radar in Japan (the MU radar). Marked changes of vertical velocity and horizontal wind direction were found between the inside and outside of the cold vortex. The vertical velocity activity outside the vortex was asymmetric; it was most active in a sector before the vortex. Unsaturated internal gravity waves in their generation stage contribute predominantly to the vertical velocity activity, suggesting that tropospheric occluded cyclones may be a possible source of middle-atmospheric gravity waves through the geostrophic adjustment process.     

Internal gravity wave selection in the upper troposphere and lower stratosphere observed by the MU radar: Preliminary results

Marked wavelike variations of the lower stratospheric wind observed on 7–10 May, 1985 by an MST radar in Japan (by the MU radar) are analyzed assuming that they are induced by monochromatic internal inertio-gravity waves. These variations are mainly composed of two modes (periods: 22 and 24 hours), both of which have zonal phase velocities (C _X ) slower than the mean westerly wind (). A statistical analysis of the zonal phase velocity shows thatC _X above andC _X below the tropopause jet stream, which is considered to be a vivid proof of wave selection due to the tropospheric mean flow and upward wave emission from the tropopause jet. A comparison between the MU radar results and routine meteorological observations leads to the conclusion that the marked waves appear when the jet stream takes a maximum wind speed.     

基于地图代数的最小生成树实现方法

本文介绍了最小生成树及其常见的算法,对比栅格算法分析了基于矢量的最小生成树算法的缺点,介绍了地图代数的距离变换和基于地图代数的距离变换图生成Voronoi图、Delaunay三角网,然后根据最小生成树MST是Delaunay三角剖分的一个子集,逐次删掉Delaunay三角网中每个三角形的最长边,从而得到最小生成树,该方法不仅适用于欧氏非障碍空间,同样也适用于障碍空间的情况,解决了以往最小生成树在障碍空间下(尤其是当障碍空间中的障碍是全形态的条件下)难以求解的问题,具有一定的理论意义。     

The Geography of Education and the Education of Geography: Agricultural Extension and the Political Ecology of Education

The spatial nature of learning is increasingly a focus of geographic inquiry. I argue that the spatiality of education, which is where formal learning occurs, has the potential to shape students’ spatial imaginaries. I analyze the role the spatiality of agronomic education plays in the historical construction of the social and physical landscape in southeastern Pará, Brazil. In opposing ways, the first Green Revolution, and agrarian social movements’ more recent agroecological Green Revolution are found to structure agronomic education and spatial imaginaries. The perspectives of agricultural extension agents trained in traditional agronomic programs are compared with teachers from an agroecological school located in an Amazonian agrarian reform settlement of Brazil's Landless Workers’ Movement. I collected these data over 17 months of ethnographic fieldwork. To analyze these data, I employ a political ecology of education perspective, which highlights how education and political economy interact to mediate relations with, access to, and contestations over natural resources. The geography of education and the education of geography exist in a complicated feedback cycle: Education is not neutral but ideologically charged and affects conceptions of productive landscapes, providing students intellectual and economic power to put their visions of landscape into effect.     

可视化空间聚类挖掘算法及系统实现

空间聚类与传统聚类方法的区别之一在于空间聚类是对空间实体的集群性进行分析,在聚类过程中需考虑模式在空间分布上的一种或几种结构特征,如模式间的远近关系、拓扑关系、方位关系、疏密关系等。然而,传统聚类算法大多忽略空间结构特征对聚类结果的影响。同时,传统数据挖掘过程往往是“黑箱”作业,用户不论感兴趣与否都只能被动地接受挖掘结果,而且结果往往是抽象的、不易理解的。本文对基于MST的可视化空间数据聚类挖掘算法进行了研究,利用Delaunav三角网和MST最小生成树使得地理实体的邻接度与其他属性数据一起参与了空间聚类处理,同时用J2EE技术开发可视化空间聚类挖掘工具,为此类应用系统的建立提供了一种实用的可行方案。     

Layered clouds in the Indian monsoon region

Contrary to the prevalent belief that tropical region is characterized by convective clouds rather than by layer clouds, we have suggested that deep convective clouds occur on meso-scale, but layer clouds occur on larger synoptic-scale with a relatively small region of deep convective clouds. Sustenance of deep convective clouds is inhibited by the presence of inertio-gravity waves, which have alternating layers of upward and downward motion in the vertical. We have also shown that inertio-gravity waves generate regions of relatively strong horizontal velocity, vertically separated by layers of relatively weak horizontal velocity. Layers of strong horizontal velocity are created by inertio-gravity wave system through convergence of vertical flux of horizontal momentum. We have also suggested that horizontal convergence/divergence of moisture flux is generated by inertio-gravity waves, giving rise to vertically alternating layers of high/low humidity, and visible or sub-visible clouds. Layers of high humidity become layers of strong radar reflectivity at frequency of 53 MHz at which MST Radar at Gadanki, near Tirupati, India, operates. These observations, more than 2,50,000 in number, for vertical grid points, spread over all the months of the year, have helped us, among other observations, to arrive at these conclusions. Further, the analysis suggests that the main source of strong MST radar reflectivity is not mechanical turbulence as is commonly believed.     

视觉认知的自适应多级岛屿群空间模式提取

针对当前电子地图显示范围以及人眼视觉分辨能力的限制,该文提出了符合视觉认知规律的自适应多级岛屿群空间模式,基于动态邻近图、最小生成树、最小面积外接矩形等概念设计了岛屿群多级空间模式提取算法。实验结果表明,该方法有效顾及了岛屿群显示的空间尺度,能够自适应地生成符合显示尺度要求的岛屿群多级空间模式,提高了空间模式识别的灵活性和有效性。