Recognition of contour line from scanned military topographic maps

This paper presents a method for contour lines recognition from scanned military topographical maps. It is a difficult process, since contour lines are always overlapped or intersected with others information, and aliasing and false colors induced by the scanning process. The proposed method overcomes these difficulties using a multi-step process. First, a segmentation feature set is developed to compensate color aliasing and false colors. Next, contour lines are extracted by a relaxation algorithm, which combine spatial and color space information to get more reliable result. Finally, the resulting binary image is thinned and then geometric properties of contour line are used to match and connect broken parts. The approach is tested successfully on a number of common-conditioned military topographic maps.