Digital Processing of Remotely Sensed Multispectral Data

Although it is well recognized that photointerpretation of the Landsat imagery can provide adequate information for mapping broad categories of land cover at a reconnaissance level, it is also true that the Landsat data in digital format (CCT's) together with numerical (computer-aided) analysis techniques can provide a great deal more information at a higher level of mapping detail with mapping units of approximately half a hectare. This paper introduces the fundamental concepts involved in numerical analysis of multispectral scanner (MSS) data. Emphasis is placed in the description of the essential steps required to conduct a multispectral classification; that is, I. Pictorial Display of the Raw Data, II. Definition of the Spectral (training) Classes, III. Classification of the Entire Study Area, IV. Pictorial and Tabular Display of the Resulting Classification, and V. Evaluation of the Classification Result.