Exploratory feature space visualization in remote sensing

Di Carlo, Walter (2000) Exploratory feature space visualization in remote sensing. (MPhil thesis), Kingston University, .


An exploratory data visualization system has been developed for satellite remote sensing. It has been designed to enable users to interact with and understand aspects of multi-dimensional remote sensing data. The system uses virtual reality technology in order to immerse the user in a graphical environment in which different views of remote sensing data sets can be visualized. The system also permits the user to interact with the data using novel virtual graphical tools which can be controlled by a data glove. Immersion in the graphical space is achieved using a head mounted display which uses a sensor to detect the user's position. The software for the system has been mainly developed using an object oriented programming language C++ together with a powerful graphics library, World ToolKit. The system integrates several different kinds of analysis tools and can generate different types of views of complex data, particularly for visualizing sub-pixel information. In this work, sub-pixel information was derived from fuzzy classification and linear spectral uri-mixing of Landsat Thematic Mapper scenes. A set of visualization techniques have been extended in order to be used in the three-dimensional environment. These include image stacks, scatter plots, parallel coordinates plots, grand t.our, hyper-corner plots, and exploratory projection pursuit. The visualization of sub-pixel information has been realized using fuzzy glyphs in which t.he transparency and colour composition is used to represent fuzzy class memberships and spectral end-member abundances. The system has been tested with data obtained from two test sites in Leicestershire, Stoughton and Knighton, which have varied land cover and significant proportions of mixed pixels.

Actions (Repository Editors)

Item Control Page Item Control Page