Airborne mss for land cover classification

Publisher Website
Google Scholar
Abstract
A basic methodology for the classification of land cover using airborne multispectral scanner (MSS) imagery is outlined. This methodology includes the four stages of: waveband selection and radiometric calibration; correction for scan angle and atmosphere; training and classification and accuracy assessment. Three possible refinements were: (1) the use of a per‐field sampling scheme; (2) the inclusion of texture in the classifier and (3) the use of two dates of imagery. Ten land covers were classified using: the basic methodology; the basic methodology with refinement 1; the basic methodology with refinements 1 and 2 and the basic methodology with refinements 1, 2 and 3. The overall levels of accuracy (Kappa index) were 65%, 82%, 85% and 87% respectively. The basic methodology and refinements 1 and 2 were shown to be suitable for land cover classification. Given the extra costs involved, the case for two dates of airborne MSS data is not a strong one and further research using optimised data sets is required before its relative utility can be evaluated.