Shape assessment of generalized building footprints

Abstract

Along with the constantly improving acquisition, processing and visualization techniques, the amount of spatial data that is available today has dramatically increased. As a consequence, various data sets for one and the same area are often available, reflecting temporal changes or representations at different resolutions. Thus, there is a rising need to compare such data sets in order to assess their quality and pick out the proper one for a specific application. In this paper, the focus lies on the shape assessment of building ground plans for which various versions are available as a result of cartographic generalization. The commonly used characteristic for measuring the deviation between the outlines of polygonal objects is the Hausdorff distance. On generalized building data, however, the metric performs less intuitive from a human point of understanding. Examples resulting from the geometric simplification of footprints are given, the shortcomings of the Hausdorff distance discussed, and an alteration of the metric motivated therewith. This new metric does not allow for a direct connection of the two points of shortest distance, but restricts it to remain inside the polygonal difference between the two footprints. Besides the shortest distance, a variation of the metric is also shown to express the complexity of shape differences.