Object-Based Slum Area Mapping using Multispectral and Textural Information

Slums are often found in big cities due to the increase in urban population. The Sustainable Development Goals (SDGs) program launched by the United Nations (UN) includes the issue of slums in the 11th goal. Improving slum conditions is a priority in many countries, including Indonesia. Jakarta Province as the largest metropolitan area in Indonesia also faces this problem. The first step to addressing the problem is to identify slums. One method that can be used for this identification is mapping using remote sensing. This research conducts slum mapping using an object-based approach and Random Forest algorithm, using multispectral information, texture, and Open Buildings V2 Polygons as parameters. Based on the research results, the classification using the object approach with the Simple Non-Iterative Clustering (SNIC) algorithm with a segmentation size of 15 pixels has the best accuracy of 0.892. Based on the modeling result obtained B11, which refers to the near-infrared band, and Normalized difference Built-up Index (NDBI) is the most important multispectral feature in the model. Meanwhile, Sum Average (SA VG) is the most important texture feature in the model and Open Buildings V2 Polygons have the greatest importance in the pixel approach but not so much in the object approach.