Abstract:
In the present time, many kinds of satellite imagery are available. They are different in spatial resolution and spectral resolution. Usually the high resolution satellite data are expensive. The satellite image fusion covers multiple techniques used to combine the geometric detail of high resolution panchromatic image and the color information of a low resolution multispectral image. It produces a final image with the highest possible spatial information content while still preserving good spectral information quality. The study area is in Min Buri district, Bangkok. Land use in this area varies from urban area and agriculture area. The first step, Use the artificial neural networks in classification of Geoeye multispectral satellite image having 4 meters ground resolution. The result is land cover map of streets, building, waterway, shrubs and perennial plant. Next step, do the image fusion between Thaichote multispectral data having 15 meters ground resolution and Thaichote panchromatic data having 2 meters ground resolution using 5 techniques mentioned above. After that we put all the fused image data in artificial neural networks classification process. The overall accuracy of classified data are approximately 80%. We obtain 5 land cover maps difference. Finally compare land cover map from Thaichote fused image and land cover map from Geoeye image. The result gives matching areas corresponding approximately 50%. The change of the digital number (DN) of a pixel in fused image process of Thaichote data and DN vary from low to high of Geoeye data. This can affect the accuracy of image classification.