Vectorization of linear features on color scanned map

Cartographic feature vectorization is one of the most salient problems in the acquisition of geographic information from scanned topographic maps. This paper presents a novel automatic approach to linear feature vectorization based on image domain. By the approach, the adjacent pixels with identical color are segmented self-adaptively into basic units (segments). The linear features are vectorized based on the analysis of segments' properties and topological relations, segment clustering, sequential tracking and topological relations matching. Experimental results shows that the liner features can be vectorized rapidly and accurately by using the proposed approach, and the information of features is preserved complete.