Abstract

Rater maps are widely available for many areas around the globe. The road network that commonly exists in raster maps is an important source of road vector data for the areas for which road vectors are not readily available. To extract the road vector data from raster maps, the first step is to extract the pixels that represent roads from the raster maps (ie, the road raster layers in the raster maps) and then extract the road vector data from the road pixels. Since the roads usually overlap with other map features and raster maps very often contain noise introduced during the processes of image compression and scanning, the extraction of road pixels from raster maps is difficult. Moreover, for extracting the road vectors from road pixels, previous work commonly uses the thinning operator [6] or line grouping and parallel-line matching techniques [1] to obtain the skeletons of the connected objects composed of road pixels. The thinning operator is robust and requires no parameter tunings; however, the thinned lines (ie, the skeletons) are usually distorted around the line intersections and hence the extracted road vectors are not accurate, especially when the thinning operator is applied on thick lines. On the other hand, the line grouping and parallel-line matching techniques require manual settings on various parameters to identify the skeletons, such as the maximum difference between the slopes of two line segments to be merged.
In this paper, we present an approach for extracting accurate road vector data from raster maps with no parameter tunings. We first utilize our previous techniques to extract the road pixels from raster maps [4] and generate a …

Date

March 6, 2026

Authors

Yao-Yi Chiang, Craig A Knoblock

Journal

Proc. GREC

Volume

2009