Research Group Geoinformation

Home | Research | Publications | Download | Projects | Teaching | Staff

... to the Cartography Group

Nothegger Clemens

Automated Extraction of Landmarks


In (Raubal and Winter 2002) a formal model is proposed that allows to compute a measure of landmarkness for features of a city. The model is based on a classification of (Sorrows and Hirtle 1999) who distinguish between visual, structural, and semantic aspects of landmarkness.

Visual attributes determined in the model comprise the façade area, façade aspect ratio, deviation from bounding rectangle, color and visibility. Structural attraction is determined mainly by the road network. Used measures are the grade of connectivity of a node, and a measure for barrierness. Semantic attraction – cultural, historial or economic importance - is determined by linking external directories, for example the Kulturgüterkataster or the Yellow Pages. The individual measures are then weighted and combined to a global measure of landmarkness.

With the given model it is possible for the first time to select features from multipurpose data sets for their landmarkness automatically. These could be used to enrich wayfinding instructions with these selected features, inspired by human communication of routes (Denis et al. 1999; Michon and Denis 2001), to improve the usability of such instructions.

This thesis shall investigate the model. Specific questions to be solved are whether the proposed model is robust to variation in the parameters, whether the measures can be computed automatically, and, of course, if the results are plausible.


The following hypothesis shall be proved: The proposed model yields plausible selections sufficiently close to human perception of reality. All individual measures are readily computable and plausible by themselves.


This thesis shall investigate the usefulness of the approach described above. The test area is the first district of Vienna, an area that is highly frequented by tourists, contains many points of interest, and is in most parts pedestrian zone. Demand for pedestrian navigation support is high.

Input data for the test area has to be collected from different sources, and partly to be supplemented or surveyed. In particular the digital ‘Multipurpose Map’ produced by the city of Vienna shall be used to derive the route network and the visibility measure.

Rectified photos of façades are used to compute measures like façade area, aspect ratio, and color.

It shall be explored if it is feasible that the calculation is embedded in an automated workflow.

To evaluate the model human subject testing is used. The subjects are shown 360° panoramic views of intersections and are told to identify the landmark they would use for guiding pedestrians. If there is a significant agreement between the automatically selected landmarks and the human subjects the hypothesis shall be considered proved.


Denis, M.; Pazzaglia, F.; Cornoldi, C.; Bertolo, L., 1999: Spatial Discourse and Navigation: An Analysis of Route Directions in the City of Venice. Applied Cognitive Psychology, 13: 145-174.

Michon, P.-E.; Denis, M., 2001: When and Why are Visual Landmarks Used in Giving Directions? In: Montello, D.R. (Ed.), Spatial Information Theory. Lecture Notes in Computer Science, 2205. Springer, Berlin, pp. 292-305.

Raubal, M.; Winter, S., 2002: Enriching Wayfinding Instructions with Local Landmarks. Technical Report, Institute for Geoinformation, TU Vienna.

Sorrows, M.E.; Hirtle, S.C., 1999: The Nature of Landmarks for Real and Electronic Spaces. In: Freksa, C.; Mark, D.M. (Eds.), Spatial Information Theory. Lecture Notes in Computer Science, 1661. Springer, Berlin, pp. 37-50.

Powered by CMSimple