Satellite Image Interpretation using Spatial Reasoning

The updating of maps using remotely sensed data is an important application in the area of satellite data processing and interpretation . Maps are digitally stored in Geographic Information Systems (GIS) and are simultaneously viewed with the image to allow an interactive update. As maps are highly abstracted representations of the real world, it is necessary to find an internal representation of map and image contents to allow an automatic comparison of the two products, reconciling the quantitative differences introduced by abstraction. This paper discusses a qualitative approach to representing map and image content in addition to the quantitative information (i.e. a two level-approach). The main focus of the paper is the map-guided image interpretation process which is a necessary part of the updating process. 1.Introduction In the area of satellite data processing and interpretation, an important application is the updating of maps using remotely sensed data (i.e. satellite images). At the moment this is mainly achieved manually by human interpreters visually comparing existing cartographic maps with a satellite image of the same area. The human interpreters bring to the problem all their prior experience in image interpretation and map making, as well as background and contextual information relevant to the area of interest. Differences in the two products are visually reconciled and any changes included in the map product. Photogrammetric workstations today permit simultaneous views of the products and interactive updates of maps. What seems an easy task for human interpreters, namely comparing maps and satellite images with each other via perceiving their contents, is a non-trivial task for computers. Further, while humans can easily perform this task on small amounts of data, their performance decreases with increasing amount of data which has to be analysed and the increasing complexity of the data sets that they are exposed to. This is a strong motivation to automate the update process, considering the huge amounts of remotely sensed data recorded daily which has to be analysed in comparatively short periods of time. There have been some attempts to automate the updating of maps (i.e. GIS data), in particular road updates. Models have been used to help the update process. In Berry (1995), models are described as representations of reality in symbolic form (i.e. abstract representation). For GIS, a spatial model is the expression of mathematical relationships among mapped variables. This GIS spatial model is often used in satellite or aerial image interpretation to describe the positioning of objects (e.g. primary objects such as lines or more complex objects such as road junctions) or their relationships to each other (e.g. the relationship between two lines which characterise a road) and is required for an object to belong to a certain object class. Map Image Interpretation Interpretation Belief and Spatial Reasoning Similarity Computation Probability Computation Map Aerial Image Abstract M ap Representation Abstract Image RepresentationM ap Representation Abstract Image Representation