Building knowledge-based systems for detecting man-made structures from remotely sensed imagery

In this paper we review and discuss several emerging themes in the area of image-interpretation for automated cartography and digital mapping. Our primary observations are that the detection, interpretation, and analysis of man-made structures by remotely sensed imagery requires the integration of spatial knowledge with image-analysis and interpretation techniques. General knowledge of structural and spatial layouts for man-made structures, as well as a priori map knowledge, can be expected to constrain search during image analysis. Spatial databases consisting of geographically referenced facts should form the basis of a priori knowledge necessary to perform interpretation and have a dual role as the repository of partially compiled facts acquired during scene-analysis.