Semi-automatic extraction and vectorisation of multicoloured cartographic objects

Abstract

We describe a semi-automatic cartographic image interpretation system in which techniques previously applied to monochrome images are generalised to colour and extended to support digitization of the full range of multi-coloured cartographic objects. A snapping tool eases the selection of seed points, speeding interpretation. Line and area objects are identified automatically and a suitable target representation (skeleton or contour) produced for each. This simplifies interaction as the user does not have to vary his/her behaviour with object type. Multi-coloured objects can be constructed from their components with ease and correction tools such as snapping and undo both save time and decrease error rates.