PCA-SVM for Effective Land Cover Mapping in Bogor Area

This paper discusses the mapping of land cover in the Bogor area and classify each pixels from the Landsat satellite images as non-impervious, partial green, or impervious area. The Bogor region has important roles as a water catchment area for the surrounding cities. Lately, the area has experienced urbanization, including the increases in population. This phe-nomenon, in turn, increases the area of impervious surface. The mapping aims to provide mapping information using the PCA-SVM method from 6 spectral channels of Landsat satellite. From the experiments, the PCA-SVM with 2 principal components using Gaussian kernel and regularization parameter $C$ = 100 has the best performance. The evaluation using visual comparison of the mapping with Google Earth Pro is also conducted with satisfactory result.