Published on: IEEE Computer, July 1992, pp. 25 to 33
Creating a cartographic database often involves the acquisition of huge amounts of data from paper drawings. The acquisition is usually performed with hand-operated digitizing tables, following procedures that are time consuming, costly and error prone. Effective techniques for the automatic input of drawings into a database are difficult to implement, and the many efforts in this direction over the past 20 years have found limited success. Only recently have substantial advances been achieved in this field.
We have developed a system for the automatic acquisition of land register maps. The system converts paper-based documents for the Italian Land Register Authority into digital form for integration into an existing database. Compliance with Land Register Authority standards was a basic design issue. These standards are strictly prescribed, and geometric entities (for example, areas relevant from the viewpoint of land taxation) must be computed within narrow tolerances.
Processing of maps begins with its digitization by a scanning device. A key step in our system is the conversion from raster format to graph representation, a special binary image format suitable for processing line structures. Subsequent steps include vectorization of the line structures and recognition of the symbols interspersed in the drawing. The system requires operator interaction only to resolve ambiguities and correct errors in the automatic process. The final result is a set of descriptors for all detected map entities, which is then stored in a database.
The system relies upon, and is driven by, the semantic of land register maps. However the image representation, vectorization techniques, and optical character recognition subsystem are quite general. The methodology implemented in the system can be generalized to the acquisition of other classes of line drawings
[ ESO | VLT software | G.Chiozzi's Home Page ]
Last modified: Thu May 21 22:12:09 METDST 1998 ;