An automatic detection method for high density slums based on regularity pattern of housing using Gabor filter and GINI index

This paper presents a development of a new approach for detecting slums area, when the density of area is very high. The basic idea of this method is based on regularity pattern of housing. We explore Gabor filter and GLCP based feature extraction to obtain the regularity feature. Then, we employ GINI index decision tree for detection. The images from Google Earth were then used in the experiment to assess our method. We select the slum areas which are defined by the local government, based on the datasheet from Biro Pusat Statistik (BPS) - Indonesia Center Bureau of Statistics as the ground truth. Finally we found that our method can perform automatic detection for area that is a slum or potentially becomes a slum, based on the given satellite image.