基于主张量的时空数据特征驱动可视化方法

Principal Tensor-based Spatial-Temporal Structure Linkage Visualization

面向时空场数据高维分析与表达的需求,构建基于主张量的时空数据结构分析与动态表达模型.简述了张量的定义、基本算子及主张量分解方法,给出了基于主张量分解的多维度特征分析流程.利用张量的多维融合特性进行多维时空数据的组织与表达,设计了多维时空数据统一组织与存储方法.利用主张量分解方法,实现了时空数据不同维度结构特征的解析与动态重构,进而建立了基于主张量的多维时空数据多维度解析模型与特征驱动的时空数据联动可视化策略.以赤道太平洋海域卫星测高SSHA(Sea Surface Height Abnormity)网格数据进行实验验证.实现了基于张量的多维透视、子集提取、等值面绘制与时空体可视化等功能.利用主张量分解实现了对ENSO(El Nio-Southern Oscillation)事件时间型与空间型的解析与提取,并实现了时间、经度、纬度系数驱动下的联动可视化.实例验证表明,该方法较好再现了ENSO事件的时空分布格局与动态演化特征,并可实现对ENSO时空演化过程的多维度透视.

Focused with the needs of expression and analysis of high dimensional spatio-temporal fields, a spatio-temporal structure analysis and dynamical expression model was constructed based on Principal Tensor analysis. This paper described the definitions, basic operators and the principal tensor decomposing algorithm as well as the procedure of spatio-temporal structure parsing and dynamical expression algorithms for multidimensional feature analysis. The unified organization and storage of multidimensional spatial-temporal data mechanism was constructed based on the multidimensional unified characteristic of tensor structures. And the structure feature parsing and dynamical reconstruction of multidimensional spatio-temporal data were implemented with the help of the principal tensor decomposing method. Then a feature-driven multidimensional-linked visualization mechanism was designed to construct the framework that integrated the processes of multidimensional data representation, characteristics extraction and linkage visualization of spatial temporal field data. The method was verified with the Grid data of Equatorial Pacific Ocean sea surface height abnormity （SSHA）acquired bysatellite altimetry, including such operations as multidimensional perspective from different dimensions, sub-data extractions, isolation surface drawing and spatio-temporal volume visualization. The temporal and spatial patterns of the El Nino-Southern Oscillation （ENSO） event was extracted and reconstructed by the Principal Tensor decomposing and linked visualization were formed with the structures force from the temporal the latitude and the longitude coefficiences. The experiment results suggested that the spatio-temporal structure analysis and dynamical expression model based on the Principal Tensor could well reproduce the spatial-temporal distribution and dynamical evolution characteristics of ENSO events. Our model could effectively support organization, storage, expression and analysis of four-dimensional spatial-temporal data and will provide good promotion and reference for multi-dimensional spatial-temporal analysis.