Knowledge-based updating of maps by interpretation of aerial images

The aim of this paper is to present our approach for interpretation of aerial images in the context of map updating. We describe concepts for the design of a top-down process that generates goals for detection of changed and new objects. The complexity of aerial photographs requires the use of context. We show that contextual reasoning provides a mechanism to generate goals to search objects by their optional or essential spatial relationship with objects already recognized. Each goal operates on a restricted search area and uses an image processing technique with parameters tailored to the searched object type. Furthermore, a priori knowledge about adaptation of the search area and the parameter setting for image processing is provided. The outcome of each goal is matched with a priori knowledge about properties of the searched object type. The concepts are illustrated by a case study for updating of a road map