Object-oriented land cover classification of panchromatic KOMPSAT-1 and SPOT-5 data

Abstract

In this paper the use of an object-oriented classification approach for the identification of basic land cover types from panchromatic images is examined. Based on data recorded by KOMPSAT-1 and SPOT-5 the same classification procedure will be applied to both data sets and the results compared to a visual classification. Results show that a classifier which not only relies on the spectral values of individual pixels but rather on a number of features, calculated for previously identified image segments can be readily used for the semi-automatic classification of basic land cover types from panchromatic data. It is also robust enough to be able to analyze new data with minimal input by the user.