The role of context and model in urban aerial image interpretation focusing on buildings

Guided by a building concept model, which interprets building into different levels and scales, this paper presents a method to extract buildings in monocular urban aerial images without priori illuminating or orientation knowledge. By using a shadow context model, a method was proposed to estimate the direction of shadow cast to verify the raw segmentations. Building extractions are refined by context, and a method of partial snake with the aid of the shadow cast direction is proposed, which can sharply reduce the iteration complexity and the influence caused by illumination. The extraction of self-shadow on gable roof with a proposed mathematical roof model is also discussed in this paper.