The Use of Hausdorff Distance as Similarity Measure in Automatic Systems for Cartographic Update

The automatic process of reviewing and updating urban maps is still a challenge to developers of high productivity computer systems. One of the intrinsic issues of this process is to quantify the similarity between objects represented in the vector space (digital cartographic bases) and that segmented in the spatial matrix (high resolution images). The research developed in the Photogrammetry Laboratory of the Department of Geomatics (UFPR), presented here, assumes the existence of qualified aerial images, which provide the adequate segmentation and the association of these segments to the generated buildings, and the adequate georeference of both data sets. From this result, the proposed method exploits a robust topological metrics, the Hausdorff Distance, to determine the up to date condition of the cartographic representation of elements in the image. Performance tests of this metric shows a high rate of agreement when compared with professional photo interpretation, and has shown that the measure is robust to partial obstruction of the compared objects

Keywords:
Hausdorff Distance; Digital Imagem Processing; Similarity Measure; Automated Cartography