As part of an effort to speed up digital vector map production for use in military geographic information systems, SRI has developed methods for automatically extracting information from raster-scanned maps. The extracted information can then be used to locate, identify, and attribute features in corresponding aerial or satellite imagery.

Complex color topographic maps contain several layers of information that overlap substantially, often within a single color plane.

Using robust techniques similar to those in SRI's location of text addresses on mail pieces and key-word spotting, SRI demonstrated that text labels on the maps can be found even amid interfering graphics. After roads in the raster map were extracted and recognized, logical links between the roads and text labels were hypothesized, permitting the names of the roads to be deduced. We have also applied our model-based recognition approaches to locate and recognize map symbols (such as the marsh symbols in USGS topographic maps) that touch and overlap other map information. (see figure below)