Research On Visualization Of Multivariate Spatial Data Field Based On Attribute Association

Visualization technology is an important means to showing the data characteristics and performing the data correlation.Multivariate space data as a representative in the field of scientific visualization,how to excavate the inner relation between variables is emphasis in the multivariate space data visualization field.Revealing the inner law of data and displaying data characteristics are research focus in the visualization.This paper from two aspects: visual rendering method and visual interaction,puts forward a visual method.The visual method for multivariate spatial data is based on attribute association.The method implements simple,convenient and interactive visualization for multivariate space data.Firstly,in order to solve the problem that it is difficult to design transfer function for multivariate spatial data visualization,we improve the traditional parallel coordinate for Multivariate space data.This paper puts forward a master variable selection algorithm based on attribute association.Combining parallel coordinate with one-dimensional transfer function curve and applying them to multivariate spatial data visualization.Master variable is the domain of the one-dimensional transfer function.The approach makes full use of the advantage of parallel coordinates data analysis and interaction of a one-dimensional transmission function.Secondly,a raycasting algorithm based on GPU is designed.The mulvariate spatial data stores in GPU in the form of 3D texture,The algorithm makes full use of GPU paraller processing ability to compute lights integral.Using parallel computing and light early termination technique to improve the rendering speed.Thirdly,according to traditional parallel coordinates' weakness that the data in the parallel coordinates is chaotic,this paper puts forward a new parallel coordinates.Multivariate spatial data is shown in the parallel coordinates in the form of strip.Combining heat map with parallel coordinates to display data distribution of each variable.This paper also designs a semantics transfer function to quickly positioning feature area of the multivariate spatial data.Finally,a multivariate spatial data visualization schema which completes the above mentioned methods is designed,conducting the experiment with the data of hurricane simulation.The experiment proves the validity of those methods put forward by this paper.