Multi-variate scientific data visualization and analytics
Abstract
Scientific computational simulations are increasing rapidly in capability and scale, producing massive amount of data that must to be processed and analyzed. The data products are often volumetric datasets in multi-variate, multi-scale, and multi-source format, exacerbating the tasks of effectively exploring, analyzing and conveying scientific information. However, traditional analysis tools often fail to provide sufficient visualization and analytics abilities to help scientists analyze and understand the simulations. An integrated analytics environment that combines visualizations, interactive navigation, comparative operators, and querying capabilities is needed. Motivated by this requirement, we developed a multi-variate scientific data visualization and analytics framework for simultaneously performing spatial, structural, and numerical analysis in an integrated environment. This new framework enables users sufficient flexibility on selecting, processing, filtering and combining of multi-variate and multi-source datasets for exploration. Statistical and numerical computation modules provide users insight of the datasets for examination and validation. Feature computation, classification and separation are performed in these modules upon users’ interests and the processed information is reflected in the transfer function design stage. Users are able to design standard 1D, 2D, n-dimensional and feature enhanced transfer functions based on numerical and statistical quantities to visually explore the datasets. Various rendering techniques, including photorealistic rendering, illustrative rendering as well as experimental photography inspired rendering, are utilized under the framework so that users can select rendering styles freely to achieve feature enhancement and separation purposes. Users are within our integrated analysis environment formed by the data analysis and visualization components so that they are able to process, visualize and obtain feedback instantaneously. The main contributions of this thesis are the following: (1) Design of a new multi-variate data visualization and analytics environment framework for simultaneously performing spatial, structural, and numerical analysis in an integrated environment; (2) Implementation of a non-uniform structured volume data rendering technique based upon rectilinear volume rendering system with multi-source data visualization abilities; (3) Implementation of a novel pre-integrated Projected Tetrahedra (PT) rendering algorithm that is capable of multi-variate transfer function design and feature enhancements of tetrahedral grid data; (4) Development of a set of feature-oriented visualization techniques that operate explicitly on tetrahedral grids, including Schlieren, shadowgraphy, silhouette and contour rendering; (5) Development of a scatterdice interface that allows users to interactively explore the multi-dimensional feature space of a volume and sculpt transfer functions across dimensions; (6) Development of data visualization tools guided by our new visualization and analytics framework to solve problems in meteorological research and computational fluid dynamics simulations. The implementations of our meteorological data and Computational Fluid Dynamics (CFD) data visualization and analytics systems are guided by our integrated multi-variate scientific data visualization and analytics framework. Case studies using these systems show that the features can be efficiently identified, numerically validated and visualized within the environments in an easy-to-use, effective manner.
Degree
Ph.D.
Advisors
Ebert, Purdue University.
Subject Area
Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.