Intelligent Systems Laboratory
Hyperspectral and Sub-pixel Remote Sensing
The goal of this project is to develop detection algorithms by which
materials which only cover a fraction of a pixel can be detected in
multispectral and hyperspectral imagery. The techniques employed involve
hypothesizing a mixing fraction and then performing a spatial
normalization of each pixel with each of its neighbors using the
hypothesized mixing fraction. Parametric classifiers and non-parametric
decision trees will be used for the purpose of performing the classification.
