THE POTENTIAL FOR VISUAL AND AUTOMATED INTERPRETATION OF FOREST VEGETATION WITH SCANNER IMAGERY

Feature classification maps derived from visual and automated methods of interpreting band-specific and composite imagery from the “Fragment” multispectral scanning system are compared in the study of vegetation and related features along the Gulf of Riga. The automated method, featuring a two-stage unsupervised/supervised classification algorithm developed at Moscow University (see MSRS, 1984, No. 3, pp. 255-261) provided for enhanced discrimination of wetland areas, farm land, and settlements, as well as for the elimination of extraneous components (especially the above) visually classified as deciduous forest. Translated from: Vestnilk Moskovskogo Universiteta, geografiya, 1988, No. 3, pp. 49-57 by Jay K. Mitchell, PlanEcon, Inc., Washington, DC 20005.