Image-Based Reconstruction of Informal Settlements

The improvement of living conditions in informal settlements (IS) is one of the most complex and pressing challenges facing developing countries. In this paper we report on progress towards the automated generation of geospatial databases of informal settlements from large scale, low cost, digital still-video imagery. We focus on shack extraction as the predominant data requirement. In contrast to western residential environments where many structures are geometrically complex yet texturally regular, sample studies in informal settlements suggest that up to 90% of shacks are approximately rectangular in shape with a flat roof, but texturally complex due to the diverse construction materials. We present a strategy for automating the detection and extraction of these structures. Preliminary results for different cues are discussed. Central to this strategy is the integration of cues in map (object) space. First experiences suggest a framework based on attributed contours, 2.5D blobs derived from digital surface models and extracted shadows will suffice for modeling shack outlines plus a representative height.