Eine Abfragesprache für die Geometrie von Rasterelementen für die rasterorientierte kartographische Mustererkennung und Datenanalyse

This dissertation deals with new possibilities for the raster-based processing of topographic maps and presents raster-oriented approaches for the structuring, management, and analysis of the contents of raster maps. The research lies at the intersection of themes from geo-information systems, cartography, and pattern recognition and concerns aspects of data acquisition, the use of pattern recognition, and data analysis in raster-based geo-information systems. The main areas of investigation are the segmentation and computation of the features of raster elements and the development of a query language for their geometry. The segmentation process isolates and encodes single coherent areas in a binary raster image and is based upon a run length encoded data structure, which allows simple management of and access to isolated raster elements. Geometrical features, such as area, perimeter, and coordinates of the centre of gravity, are computed for every raster element and stored in a multidimensional feature vector. A query language, similar to the well-known database language „SQL“ and enhanced by the addition of newly developed language elements, facilitates access to the values of the computed features. This allows an interactive and intuitive classification and analysis of the content of single colour layers in the Swiss topographic map. The new language elements permit the following: the re-use of the results of a previous query the visual display of the query result the processing of so-called complex queries, which are collections of multiple single queries The application of the query language proves to be a tool suitable for pattern recognition and data analysis. The clear extent of the query language and its similarity to „SQL“ allow the user to quickly formulate desired queries. The re-use of a query result is important for pattern recognition and allows the verification of intermediate results of simple queries. The simple queries are then combined and used as a complex query which is stored as a keyword for a pattern. The segmentation process, the computation of features, and the query functionality are implemented in the prototype SoftwareRaQueL (RasterQueryLanguage). The results obtained show that structured raster data makes possible the direct recognition of patterns and direct analysis of data from the original raster image, as well as the combination of familiar digital image processing methods with the new segmentation techniques and query language. The approach presented is suitable for data acquisition, as well as for the possible management of structured raster data in geo-information systems. From this we may conclude that more intensive efforts should be made to study the potential of structuring and analyzing raster data, which might allow the realization of genuinely hybrid geo-information systems that can manage both vector and raster data at the same time.