Download (PDF 283 KB): https://research.fs.usda.gov/download/treesearch/59374.pdf
Abstract
Optimum channel selection among 12 channels of multispectral scanner imagery identified six as providing the best information about 11 vegetation classes and two nonvegetation classes at the Manitou Experimental Forest (NASA Test Site 242). Intensive pre-processing of the scanner signals was required to eliminate a serious scan angle effect. Final processing of the normalized data provided acceptable recognition results of generalized plant community types. Serious errors occurred with attempts to classify specific community types within upland grassland areas. The consideration of the convex mixtures concept--effects of amounts of live plant cover, exposed soil, and plant litter cover on apparent scene radiances-- significantly improved the classification of some of the grassland classes. The data processed was obtained as part of Mission 19 at 1000 hours on July 29, 1970, by the University of Michigan multispectral scanner flown at 915 meters (3,000 feet) above mean terrain elevation.
Citation
Driscoll, Richard S. 1972. Pattern recognition of native plant communities - Manitou Colorado test site. In: 4th Annual Earth Resources Program Review: Volume 5: Agriculture and Forestry Programs. p. 123-1-123-12.