Band-Moment Analysis of Imaging- Spectrometer Data

Modern sensor systems high in spectral resolution, such as the Airborne Imaging Spectrometer (AIS), are challenging from an image-processing standpoint because direct classification of datasets with 100 or more spectral channels is virtually impossible. Thus, it is necessary to conduct pre-classification manipulations and transformations both to reduce the amount of data to be classified and to extract pertinent information from the original dataset. Bandmoment analysis is a simple, computationally efficient, and productive alternative to principal-components analysis (PCA), a technique commonly used for data reduction. Results from moment analysis of AIS-l data for western Nebraska are compared to those derived from PCA. Although additional work is warranted, especially concerning the physical meaning of individual moments, several advantages of moment analysis are documented including computational efficiency, reduction of sensor noise, and an overall image quality which is at least as good as a first-principal-component image.