Terrain classification in urban wetlands with high-spatial resolution multi-spectral imagery

The Elkhorn Slough area is a major wildlife reserve, and important wetlands area along the central coast of California. As part of the ongoing study of the slough, accurate classifications of the regions in the slough and its watershed are needed. The objective of this work was to use the 4-m spatial resolution, 4-color sensor on the IKONOS satellite. A manually constructed classification map was used to train the spectral classifiers. As the work evolved, a problem emerged due to the relatively high spatial resolution. Regions of interest such as oak woodlands have highly variable spectral elements when observed at high spatial resolution. As a consequence, the refinement of the regions of interest obtained from the original classification map required not only the elimination of erroneous spectral elements; but also required inclusion of a range of spectra, as opposed to the traditional approach of selecting pure exemplars. Modest success was obtained from the classification effort. As a consequence, additional virtual bands were created by constructing texture features from the higher spatial resolution panchromatic band. This enabled spectrally similar classes such as trees and cultivated fields to be distinguished.