Abstract
A new elevation labeling method is proposed which can be applied to contour images including broken lines and inconsistencies caused by loss of contour lines and/or existence of irrelavant lines. Because conventional elevation labeling methods need ideal contours as an input data, broken lines and inconsistencies must be eliminated from the target contour image before elevation labeling. This elimination process (editing) has been done by human operator conventionally. The time-consuming human operation is a large problem in DEM generation from topographical maps. The proposed method has been developed In order to remove the above human operation. In the proposed method, a probablistic relaxation approach is used. That is, elevation probabilities are assigned to each contour, and then the probabilities are propagated to neighbor contours so as to decrease local conflicts.As a result, consistent labeling can be achieved even if the target contour image may include broken lines and some inconsistencies.
