Determining the dimensionality of hyperspectral imagery

Hyperspectral systems have significantly progressed through recent advancements in sensor technology. Hyperspectal images have differents applications in geology. High dimensionality of data sets makes it difficult to apply statistical models to full images. For this reason, it is necessary to apply methods for data reduction. The use of more dimensions than strictly necessary leads to several problems such as storage and processing, therefore it is necessary to determine the dimensionality of the data. These thesis presents different linear and nonlinear methods for the estimation of the dimensionality of hyperspectral images. This methods are reviewed and applied to synthetic and AVIRIS data. The performance of the determination the of linear dimension is studied using supervised classification in hyperspectral data.