Encapsulating And Preprocessing Datasets In Scientific Visualization

Datasets generated by scientific numerical simulation give a great diversity of types. These can be characterized according to several rules. First, diferent discretion method in scientific computational problem leads to various kinds of grid, such as regular, structured, unstructured and hybrid grid. Second, attribute data can be associated with dataset points or cells, but sometimes attribute data may be assigned in scattered mode. Third, attribute data is often categorized into specific types of data. These categories have been created in response to common data forms, such as scalars, vectors and tensors. Sometimes a few attribute datas , such as tempreture, density and pressure, are associated with a single grid. Finally, many a storing format for dataset is available. Storing in parallel appling to great scale datasets varies from layering to blocking. Above-mentioned can be categorized into grid carrier, geometry and topology structure, physics attribute and storing method for datasets. Categorizing and preprocesing all kinds of dataset above are exceedingly crucial for visualization.In object-oriented design and implementation of visualiztion system it is very important to build uniform framework for object. It is also critical to encapsulate and preprocess diverse datasets so they can be integrated into processing pipeline in visualization. We give a research on many effective technologies on encapsulating and preprocessing all datasets based on our two dimentional general serial visualization system 2D-DVS. We also make a advanced study on preprocessing great scale datasets which is beneficial to developing distributed and parallel visualization system popular today.The main contributions of the paper can be classified as the following:(1) Classify datasets.We give several calssification way for datasets according to grid carrier, geometry and topology structure, physics attribute and storing method. And We build a varity of datasets classes based on object-oriented design.(2) Encapsulate datasets. We have made much research on technologies of encapsulating dataset readers on 2D-DVS.(3) Encapsulate preprocessing procedure for datasets.We encapsulate filtering procedure into different process class.Different filtering procedure lead to different visualization results.We encapsulate filtering procedure into over thirty process classes according to various visualization methods.(4) Preprocess large datasets.We reorganized the structure of great scale datasets for preprocessing. We also study the grid partition method and hiberarchy structure in the newest version of VTK. Meta-cell and scalar tree are introduced to accelerating visualizing datasets. Technologies above have been applied to dealing with the real datasets from the computational numerical simulation.Some experience and innovation on developing visualization system is summarized at the end of the paper. Some future research direction are bringed forward after that.