An Evaluation of Geometric Data Acquisition Using Landsat Imagery.

Abstract

The implementation of appropriate digital image processing method is crucial for deriving urban land cover maps of acceptable accuracy and cost. This study examines the effect of acquiring images in various spectral regions (bands), the impact of some image processing techniques on the combination of the different bands and the acceptable mode in which the features of the image could be classified using unsupervised classification (clustering) and supervised classification based on four different hard classifiers. Four different filter types were experimented on the colour composite images before classifying the images into different distinct land spectral classes. The Integrated Land and Water Information System (ILWIS) software was used to classify LandSAT 7 image of 2001, part 189r053, zone 32, bands 1 (Blue), 2 (Green), 3 (Red), 4 (Near infrared), 5 and 7 (Middle infrared) wavelength. From the study, it was observed that AVG 3x3 filter type is the most preferred. Colour composite of bands 5, 4, 3 in the RGB planes gave the best representation of the features of the image and that Box classifier, Minimum Distance to Mean Classifier and Maximum Likelihood classifier are excellent classifiers for image supervised classification.