Multi-mode Tensor Expression Model of Multidimensional Spatio-temporal Field Data
HU Yon · Wuhan Daxue xuebao. Xinxi kexue ban · 2015
The effective management and analysis of multidimensional spatio-temporal field data with increased quantity and complexity is a bottleneck in current GIS.This paper introduces a tensor structure with multidimensional unity and coordinate independence to provide a multi-mode tensor expression model for spatio-temporal field data.We realize tensor construction for geosciences data and design data and task flows for multidimensional tensor data analysis.The management and analysis of multidimensional spatio-temporal field data is aimed at different application contexts;realized in the paper by defining three kinds of tensor data expression and application models based on original tensor,tensor decomposition and hierarchical tensor decomposition,respectively;according to different data management and analysis needs.On this basis,we designed the tensor-based multi-mode data analysis framework and application analysis traffic flow template.We implemented these key technologies,integrated them in a prototype system,and tested its performance based on the satellite altimetry measurements.The results show that this new type of tensor-based multi-mode expression model can support the management,analysis and storage of massive spatio-temporal field data.It also has an advantage for memory space occupancy and the efficiency of retrieval and analysis.