Robustness of rule sets using VHR imagery to detect informal settlements : a case of Mumbai, India

Abstract

Robust monitoring approaches for informal settlements using very high-resolution (VHR) satellite imagery can deliver essential information for supporting the formulation of pro-poor policies. Such information can complement census methods or participatory approaches. With the increasing availability of VHR satellite imagery, detection of the informal settlements benefits from the conceptualization of location-specific knowledge in the form of a locally-adapted generic slum ontology (GSO). In this study, we developed the local slum ontology for Mumbai, India, by incorporating local knowledge with image-based proxies. Then, we translated the local ontology into a rule set using Object Based Image Analysis (OBIA) to identify informal settlements by using spectral, spatial, geometric and texture measures. The method was applied to three subsets of a Worldview-2 imagery. The robustness of the initial rule set was analysed with the help of membership functions. The results showed that the normalized difference ratio of near infrared (NIR) and blue band and grey level co-occurrence matrix (GLCM) features are most effective in all three subsets in the identification of informal settlements. The results suggest that the rule sets developed in this study can potentially be applied to other study areas of Worldview-2 imagery for informal settlements identification.