Recent developments in remote sensing technologies have made high resolution remotely sensed data such as hyperspectral and synthetic aperture radar (SAR) data readily available to detect and classify objects on the earth using pattern recognition. However, the dimensionality of such remotely sensed data is often large relative to the number of training samples available. Hence, dimensionality reduction technologies are often adopted to overcome the “curse of dimensionality” phenomenon. This present thesis focuses on the problem of dimensionality reduction of remote sensing data by proposing two algorithms for robust classification of hyperspectral and SAR data. Specifically, for hyperspectral image analysis, a genetic algorithm based feature selection and linear discriminant analysis based dimensionality reduction method is proposed, and, for SAR data, polarization channel based feature grouping followed by a multi-classifier, decision fusion technique is proposed. The algorithmic framework of the proposed approaches and experimental results will be presented in this thesis.
| Identifer | oai:union.ndltd.org:MSSTATE/oai:scholarsjunction.msstate.edu:td-3390 |
| Date | 12 May 2012 |
| Creators | Cui, Minshan |
| Publisher | Scholars Junction |
| Source Sets | Mississippi State University |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Theses and Dissertations |
Page generated in 0.002 seconds