A Topological Approach to Processing Scanned Urban Maps

For many GIS applications the data entry component is the most expensive, and frequently makes the difference between success and failure of the project, in terms of both time and money. The most time-consuming part of manual digitizing is usually the generation of correct topology, which usually involves many of error correction steps. The automatic processing of scanned maps appears to have many advantages, but in practice the line extraction process does not always produce good topology automatically. The situation with scanned maps may be improved with a conceptual change of emphasis. Instead of concentrating on the black pixels (linework) one may emphasize the white pixels (polygonal areas) and build the relationships from these. The process as implemented consists of three steps. Firstly a set of white pixels are selected, at a user-specified distance from the black linework, and these are given a polygon label on the basis of a flood-fill algorithm that scans all connected white pixels. Secondly these selected pixels are used as data points to generate the standard Euclidean point Voronoi diagram. Thirdly, this structure is scanned to extract Voronoi boundaries between pixels having different polygon labels. The result of this operation is a set of vector chains or arcs that are guaranteed to separate regions of white space, and are guaranteed to connect precisely at nodes. Experiments on a small cadastral map show excellent automatic topology generation, and accurate linework. Black linework or symbols that are unclosed are not preserved with this method which may be an advantage or a disadvantage, depending on the application. Unclosed polygons are also lost, and may need pretreatment. Small polygons generated as the interiors of closed symbols may be removed automatically in most cases by simple filters. The results are imported directly into the vector GIS without any cleanup or intersection-testing operations. In many cases the approach described here may make significant savings in the implementation of urban planning systems.