Comparing the accuracy of machine classified landsat imagery to manually delineated aerial photographs for county appraisal district use

Abstract

The purpose of this study was to compare the results of a computer classified satellite image to human classified aerial Photographs, and to determine the accuracy of the satellite image for land-use/land-cover mapping of forested areas. One 1:24,000 quadrangle map (Chester) was used to test the supervised and unsupervised classification methods and a second (Boggy Lake) was used to verify the classification. The resulting maps were then tested for accuracy by comparing them to human classified aerial photographs of the same area. Verification was accomplished using systematically located points with a random start. Each map was analyzed for errors of omission and errors of commission. The results of this study showed accuracy levels of the satellite images as compared to the aerial photographs were below 50 percent for both quadrangle map areas and both classification methods. No significant difference was detected between the supervised and the unsupervised classification methods for either quadrangle map area. The verification process using the Boggy Lake Quadrangle map, did not show a significant difference from the developmental map, the Chester Quadrangle, for either classification method.