FUSI CITRA LANDSAT 7 ETM+ MULTISPEKTRAL DAN PANKROMATIK UNTUK MENINGKATKAN HASIL KLASIFIKASI

Advancement of remote sensing technology is able to provide remote sensing images that has high spatial resolution, spectral resolution and temporal resolution. However, the remote sensing satellite sensor system, the spatial resolution and spectral resolution images are mutually contradictory. Some remote sensing satellite able to provide multispectral imagery with information that can distinguish features spectrally but not spatially. Image fusion or pan-sharpening is a proper technique to corporate detailed geometry (spatial) and detailed color (spectral) at the original image pair to obtain a new multispectral imagery with sharper spatial and spectral information. From the results, it provide an alternative to produce a better accuracy of land use classification with medium data images that are economical and readily available. The image used is a Landsat 7 ETM + Multispectral (composite of Band 1, 2, 3, 4, 5 and 7) and Landsat 7 ETM + panchromatic (band 8) with a spatial resolution of each image is 30 meters and 15 meters. Selected image fusion process is at pixel level using Brovey transformation method. The image produced from the fusion process then be used as input image for the land use and land cover classification by the maximum likelihood method. The reliability of image fusion results performed by statistical data analysis such as image histogram and statistical value of image constituent, while for the validation of classification results conducted by field tests, separability index analysis of training area and calculate the accuracy by confusion matrix of the classification results. From the image fusion process and use-land cover classification prove that Landsat 7 ETM + Multispectral and pancromatic image can be used as an input pixel level fusion (Brovey method). Fusion of Landsat 7 ETM + Multispectral and panchromatic Brovey using to improve the spatial resolution and the contrast of the image. But it is inversely proportional to radiometric increasement. The classification using image fusion results provide object recognition rate more than multispectral or panchromatic imagery. Based on field tests, the results of classification using image fusion has a high accuracy, seen by suitability of land use and coordinate both in image and in the field. Overall classification accuracy up to 90% with an average user accuracy and producer accuracy 91,27 % of 92,05 %