Automated Vectorization and Labeling of Very Large Raster Hypsographic Map Images Using Contour Graph

This paper presents a very efficient method for vectorizing and labeling very large raster hypsographic map images. By extending the contour tree (Freeman and Morse, 1967) to a contour graph, open contour lines and carrying contour lines can be handled. The ambiguities of the topological relationships among contour lines are resolved using map structure information, i.e., elevation tags and index contour lines. Contour graph is constructed and processed efficiently by introducing index contour graph. By using a sequential extraction algorithm (Agrawala and Kulkarni, 1977) for vectorization, attributes for contour lines, such as line width, and elevation tag(s), are extracted efficiently. Although it was developed primarily for USGS 1:24,000 hypsographic maps as part of the GISOM project at The Ohio State University Center for Mapping, the method also is applicable to other scale USGS hypsographic maps. We have applied the algorithm successfully to raster images with pixel size of 19,320 by 25,400. On average, it can correctly vectorize more than 98 percent of the contour lines and correctly label more than 90 percent.