Image segmentation is basic task of object-based image analysis. The result is the creation of image objects defined as individual areas with shape and spectral homogeneity. In practice, image analyst requires trial and error to justify an optimum scale for multiresolution segmentation. The main objective was to examine an optimum scale of multiresolution segmentation applying to pan-sharpened Landsat-8 data for land use and land cover classification under object-based image analysis (OBIA). In this study, a systematic scale of multiresolution segmentation were firstly set up at scale of 10, 15, 20, 25 and 30 and applied to pan-sharpened Landsat-8 data acquired in 2013 for creating a corresponding image objects. Then, the derived image objects by each scale were further classified LULC map using a nearest neighbor classification with feature optimization. After that the individual derived LULC data by each scale was accessed its accuracy using overall accuracy and Kappa hat coefficient based on the reference data from land use data in 2011 of Land Development Department. The scale which provides the highest overall accuracy and Kappa hat coefficient values will be chosen as an optimum scale of multiresolution segmentation. The result showed that the optimum scale of multiresolution segmentation of pan-sharpened Landsat-8 data for LULC extraction under OBIA was 30 with shape parameter at 0.1 and compactness © The 41st Congress on Science and Technology of Thailand (STT41)
parameter at 0.5. This finding can be used as a guideline to image analyst for multiresolution segmentation of pan-sharpened Landsat-8 data without trial and error.