[P1-09] Standardizing Land Cover Classification of Landsat 9 OLI through Pixel-Based ROI Expansion
Keywords:Land Cover Classification, Landsat
The present study examines the characteristics and appropriateness of the method for selecting regions of interest (ROIs) for land cover classification using two Landsat-9 OLI/TIRS satellite images of different seasons in Gumi, South Korea. The distribution of land cover in the target area can affect the classification method and results, making it difficult to accurately classify when the distribution of a specific land cover is small or distributed adjacent to others. To address this issue, we specified a base pixel for each land cover when selecting the ROI and incremented pixels to the neighboring pixels from the base. We generated a total of 30 pixels for each ROI, ranging from one to a maximum of 30 pixels. We compared and analyzed the changes in accuracies according to the number of pixels for the two seasons and investigated existing optimized pixel numbers for best land cover classification accuracy.