Development of Visual Data Mining Tool for Vector Fields

点击翻译按钮获取中文摘要

In this research, a texture-based vector visualization method, called LIC (Line Integral Convolution), has been extended to three-dimensional fields. A sophisticated vector volume visualization code has been developed, which can adaptively visualize inner structures of interest around critical points within generated 3D LIC textures by topologically-accentuated spatial transfer functions and a 3D streamline illumination model.The core algorithm consists of the following four steps :1) Generate a 3D volume for solid LIC texture from a given vector field by using a method developed by Mao, et al (1997).2) Locate critical points of the field with differential topology methods.3) Accentuate spatial transfer functions in a topological manner so that they can reflect the significance distribution analyzed in Step 2.4) Volume render the LIC volume generated in Step 1 with the spatial transfer functions and a 3D streamline illumination model.The above method has been implemented on a GPU-enhanced volume graphics PC and cluster systems for providing a real-time volume data mining environment. The effectiveness of the tool has been proved empirically with an application to several practical datasets, including a simulated tornado dataset provided by R. Crawfis of Ohio State University, US, and geo-scientific datasets generated by using GeoFEM, known as a parallel FEM middleware for the Earth Simulator.