Publications (2)0 Total impact
ABSTRACT: We present work in progress on the development of a partially-automated system for interpretation of map images. The principal aim of our project is to demonstrate that an adaptive system can decrease operator intervention with increasing conversion volume. The extracted information is evaluated against an established database. The current level of automation yields a 35% saving in operator time over manual entry. A further 30% appears achievable with known techniques.
04/2006: pages 302-313;
ABSTRACT: es that facilitate the batch conversion of maps of the same type. While the conversion of the first map in the batch may require significant operator intervention, subsequent maps will benefit from adaptive techniques based on the graphic and typesetting consistency within the batch. We build on considerable previous work. Vectorization under document-specific constraints and "beautification" are discussed in [1, 2, 3, 4], street-line and street-name extraction in [5, 6], and the separation of text and line art in [7, 8]. Excellent overviews of model-based map interpretation can be found in two recent dissertations [9, 10]. We also believe that over the last two decades much cartographic and image processing expertise has been built into our principal software tool, ARC/INFO . The data flow through the conversion process is illustrated in Figure 1. The registration of the map to geodesic coordinates is required only for evaluation against the reference data. Color separation is cur