Application of perceptual grouping concepts to the recognition of residential buildings in high resolution satellite images

A methodology is presented for the extraction of housing units from segmented renditions of high resolution (1-3 metres) monochrome satellite imagery. In this resolution regime, segments encompassing individual buildings are difficult to discern based solely on their inherent spectral and spatial attributes. An alternative recognition strategy is developed which employs perceptual grouping concepts of collinearity, similarity, proximity and continuity to support the definition of groups of houses that face the same street. Testing on imagery of a suburban landscape indicates that automated group processing is useful and cat? assist in reducing subsequent manual processing by 50 to 60%.