Abstract
A class definition method has been developed for extracting the land-cover information from multispectral scanner (MSS) data. The method is based on the statistical verification tested by F-values and t-values, which are estimated by multiple regression analysis in a stepwise manner. It has been applied to Landsat and aircraft MSS data. Based on the statistical verification, the original land-cover classes which are interpreted manually with color aerial photographs, have been consolidated in sequence. Consequently, thirteen classes for the Landsat MSS data and eleven classes for the aircraft MSS data are defined.
