Land use classification based on object and pixel using Landsat 8 OLI in Kendari City, Southeast Sulawesi Province, Indonesia

Abstract

Abstract Current development on method and technique for data processing in remote sensing has received a great importance, since it provides a base for a myriad of applications, including land use monitoring. OBIA- and pixel-based approach are commonly used as classification technique in remote sensing. This present work aimed to compare both approaches in classifying the land use in Kendari, Southeast Sulawesi Province, using Landsat 8 OLI (Operational Land Imager). Digital images were processed using both techniques with the use of support vector machine (SVM) algorithm. The 188 sampling spots in the study site were randomly determined, then classified into 5 groups of land use: water body, follow land, built up land, residential area, and vegetation. Data obtained were then validated according to data from google earth and ground check. The accuracy was assessed by confusion matrix method using Region of Interest (ROI). The results showed that OBIA-based classification coupled with SVM algorithm showed overall accuracy of 81.38%, Kappa coefficient of 78.77%, in which the accuracy was 9.57% higher than pixel-based classification. Based on this finding, OBIA was reported to produce a better performance in identifying land use in urban area.