中文摘要：

以遥感数据、数字高程数据等为代表的栅格数据获取技术的进步,以及栅格数据本身适合地学模拟的特点,使得栅格数据应用越来越广泛。当前以定量计算为主的方法难以有效支撑栅格数据分析任务,将可视化引入,充分利用人机协同优势,形成栅格数据地学可视化分析环境是一个较好的解决途径。但是,栅格数据大数据量的特征会引起属性空间可视化时的遮挡问题,分析者难以通过可视化分析环境有效识别有意义的地学模式。本研究主要针对这一问题,在现有方法基础上,提出了一种基于体绘制的层次性栅格数据地学可视化分析环境构建方法。当栅格数据集较大时,采用体绘制方法表达密度信息,避免大数据量引起的遮挡问题;在分析者通过人机交互缩小感兴趣数据集后,采用平行坐标法进行可视化,支持细节模式的发现。新方法所构建的原型系统被成功应用于从地形数据集中发现代表土壤类型的聚类模式,从而验证了方法的有效性。

英文摘要：

Developments of raster data capture technologies and demands from application fields call for advanced raster data analysis methods.Automatic algorithms often cannot well support this need due to the complexity of geographical phenomenon and limitations of algorithms themselves.Geo-visual analytics that involve human＇s visual analytical capability in data analysis attracts attention in recent years.However,Raster datasets usually have large amount of pixels,which may cause serious clotting problem in visualizing raster data in attribute space and thus it is difficult for analysts to visually detect patterns in raster datasets.The research reported here mainly focuses on this problem.Based on existing solutions and current computer graphics technologies,we propose a new volume-rendering-based hierarchical approach to construct interactive geo-visual analysis environment for raster data.In the first hierarchy,volume rendering is used to express density information instead of original pixels in attribute space to avoid clotting problem.In the second hierarchy,after analysts select relatively small-sized sub-datasets using some interaction tools,parallel coordinates plot is used to support analysts to capture detailed patterns in attribute space.On different hierarchies of this progressive visual interface,attribute space visualizations are linked with geographic space visualization to facilitate the detection of patterns with geographic meanings.Software prototype was developed based on this idea and then applied in a terrain dataset to find small clusters that may represent possible soil types in digital soil mapping.The case study shows that the proposed approach can well support the progressive detection of geographic cluster patterns that may be neglected by automatic clustering algorithms and thus demonstrates effectiveness of the proposed approach.