Integrating geographic analysis with image databases

Abstract

Land related information about the Earth's surface is commonly found in two forms: (i) map information, and (ii) satellite image data. Map information is stored as thematic layers in a geographical information system. (GIS). Often the information is collected for a specific purpose at a point in time. It can depart from the true situation on the ground due to subjective interpretation and generalisation of the data. Satellite images capture electromagnetic energy of the Earth's surface as multi-spectral images (Landsat TM, SPOT) and radar images (ERS-1, SIR-B). The images provide a good visual picture of what is on the ground but complex image processing is required to interpret features in an image scene. A great deal of research work has occurred to improve the reliability of satellite image interpretation. Especially in the area of image classification, feature recognition, and image browsing. But at some point the ability to extract intelligent information from image data is constrained without the assistance of auxiliary information. Using already interpreted map information may make further improvement in satellite image interpretation. Image interpretation can incorporate a priori knowledge about mapped information in its analysis. The analysis should not only include image classification but also combine with a rule system to reason about geographic knowledge. For instance, one would expect to find a bridge at the intersection of a road and a river.