Slum Extraction from High Resolution Satellite Data using Mathematical Morphology based approach

Abstract

ABSTRACT This paper proposes different mathematical morphology-based approaches to detect the urban slums from urban buildings using Very High Resolution (VHR) satellite images. Due to the rapid growth of urbanization activities, slum regions are the most evolving regions in South India whose locations are improper in official maps and statistics. Thus, automatic location identification of urban slums provides vital information for urban planners to formulate pro-poor policies and allocate resources properly. To detect the slums automatically from VHR satellite images, Multi Shape-Multi Size-Morphological Profile (MSh-MSi-MP) technique is proposed in this work. The development of this algorithm is motivated by modelling the urban slums using the conventional Morphological Profile (MP) which fails to model different structures in an image. On the other hand, the proposed MSh-MSi-MP is built by the sequence of morphological opening and closing profiles with different shapes and different sizes of structuring elements. Though MSh-MSi-MP yields high accuracy, some open areas in informal settlements and roofline structures of formal settlements are still classified as slums. To optimize these misclassification results, Morphological Spatial Pattern Analysis (MSPA) is used as a post-processing operation. The four different very high-resolution satellite images of Madurai city, South India acquired by WorldView-2 Sensor (1.84 m) proved the ability of the proposed approach to identify urban slums from other features by generating higher classification accuracy than any other methods.