Raster to Vector Conversion in a Map Interpretation System

This paper describes the two components: thinning and vectorization, at the front end of a Geographic Information System. The raster dataset associated with a map is very large. A small D-sized drawing gives a 120 Mpixel document. The amount of page faults encountered by a typical conventional thinning algorithm is so large that the run time is dominated by disk 110 time. The Contour Generation Thinning Algorithm based on chain codes has been adopted and the amount of page faults are reduced by a factor equal to the total number of iterations. The skeleton is vectorized by traversing the chain codes and examining the raster. A tagging scheme is used that allows the vectorization process to identify all the end points and multiply-connected nodes. INTRODUCTION sible for the preparation of input from maps and drawings. It takes the scanned data, extracts the boundaries of line segments and symbols, thins and vectorizes them, and in the case of symbols performs recognition. The result is an interpreted document which is the input to a GIs. The magnitude of the task is tremendous because of the large dataset associated with the input document. A small D sized (22"x34") drawing digitized at 400 dots per inch gives a 120 Mpixel document. The system is usually implemented on a general purpose workstation environment for the ease of its incorporation in an existing GIs. In the following sections, the thinning and vectorization processes of a document interpretation system developed by DataSpan Technology and the Alberta Research Council are described.