Raster data structures and topographic data

Abstract

The use of computers to assist in map-making has been growing for two decades; their speed of operation, large data storage capacity and flexibility of usage have been major factors in establishing many development and working computer mapping systems throughout the world. In Britain, the Ordnance Survey has supported a digital solution to the production, storage and display of large scale topographic maps since 1972. Until now, the work of the Ordnance Survey - and, indeed, most topographic map-makers in Britain who are investigating digital techniques - have adopted a vector-based strategy to digital mapping in which the data are held as a series of coordinate-points describing the lines shown on the map images. Comparatively little work has been undertaken in Britain on the use of raster-based methods of data capture and storage in which map images are resolved into arrays of small cells or picture elements by appropriately tuned scanning devices. This alternative strategy is known - from work carried out in other countries, chiefly the United States - to be suitable for some types of data manipulation, although its suitability for Ordnance Survey mapping applications is unknown. Very little investigation has been made anywhere in the world of the manipulation of raster data structures by the recently developed array processor computers; almost all existing work is restricted to the use of traditional serial machines. This thesis reports on a three year study carried out in the University of Durham to investigate the applicability of raster data processing for the work of the British national mapping organisation. In particular, it describes the distinction between vector and raster applications with geographic data and the likely characteristics of suitable raster data structures on both serial and parallel computers. A section is also included which describes the nature of scanning trials carried out on a number of commercial devices; it has thus been possible to assess not only the likely advantages and limitations of handling British large-scale map data in raster form but also its technical feasibility. The work reports on the likely volumes of data to be expected and describes parallel algorithms for operations such as polygon creation (and, indirectly, the creation of node and link vector files).