The use of obia for extraction of land cover and land use in the city of palu

Abstract

The City of Palu, a seaside urban area with active physical development, continues to face land use complexity. Since the availability of detailed data is the key to dealing with the impact of land cover change, OBIA offers an alternative to object-based image processing. Extracting land cover/land use information on the coast of Palu is a new challenge. This study used SPOT-6 recorded on September 22, 2018, that had been processed through mosaicking and cloud masking to produce a cloud-free multispectral image. Spectral channels from SPOT-6 served as an input to two research stages, namely, data segmentation and classification. The former used the Multiresolution Segmentation algorithm, while the latter applied a series of multilevel thresholds arranged into a classification-based decision tree. The extraction product, i.e., land cover/land use data, had an overall accuracy of >70%. However, there is an assumption that by performing the maximum likelihood classification technique before OBIA, the accuracy can increase. On a detailed scale, certain classes had high accuracy, including water body and vegetation. These results indicate that the combination of maximum likelihood classification and OBIA provides an alternative for identifying and extracting land cover information for coastal area mapping.