RUcore Resource Object
RUcore Resource Object
PDF-1(14.92 MB)
Uniform TitleLarge scale feature extraction and tracking
NameDhume, Pinakin (author), Silver, Deborah (chair), Wilder, Joseph (internal member), Marsic, Ivan (internal member), Rutgers University, Graduate School - New Brunswick,
Degree Date2007
Date Created2007
SubjectElectrical and Computer Engineering,
Computational grids (Computer systems),
High performance computing,
Parallel processing (Electronic computers),
Visualization
DescriptionThe goal of Scientific Visualization is to provide a more intuitive interpretation of the data being presented. Traditionally, visualization of 3D time-varying datasets is done using animation based on either iso-surfacing or volume rendering. However, for datasets with continuously evolving features it is difficult to follow and see patterns in 3D. An automated procedure to track features and to detect particular events in their evolution can help scientists to concentrate on regions and phenomena of interest.
Today, computations and simulations are performed on massively parallel computers leading to thousands of datasets where each datasets can be on the order of gigabytes. Visualizing and quantifying such data cannot be done on a single processor machine. Therefore, a distributed form of the feature extraction and feature tracking algorithms is required. In the Vizlab, we have developed a number of tools to extract and track features on a parallel cluster. However, there are cases where one would like to handle large datasets. In this thesis, we extend the feature extraction and tracking library to perform feature extraction on a large datasets with multiple steps, i.e., reading in only a portion of the data at once. In addition to the distributed feature tracking, we also enhanced the visualization component so that features can be accessed and rendered in a more intuitive fashion.
Today, computations and simulations are performed on massively parallel computers leading to thousands of datasets where each datasets can be on the order of gigabytes. Visualizing and quantifying such data cannot be done on a single processor machine. Therefore, a distributed form of the feature extraction and feature tracking algorithms is required. In the Vizlab, we have developed a number of tools to extract and track features on a parallel cluster. However, there are cases where one would like to handle large datasets. In this thesis, we extend the feature extraction and tracking library to perform feature extraction on a large datasets with multiple steps, i.e., reading in only a portion of the data at once. In addition to the distributed feature tracking, we also enhanced the visualization component so that features can be accessed and rendered in a more intuitive fashion.
NoteM.S.
NoteIncludes bibliographical references (p. 114-115).
Genretheses
Persistent URLhttp://hdl.rutgers.edu/1782.2/rucore10001600001.ETD.15832
LanguageEnglish
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.