A Novel GIS-based Polygon Shape Similarity Measure Applied to OSM Building Footprints

Abstract. Assessing the similarity between polygonal shapes is a fundamental problem in geographic information science (GIS) with applications in spatial data quality assessment, feature matching, and cartographic generalization. This paper introduces a novel and computationally efficient shape similarity measure tailored for comparing building footprints in OpenStreetMap (OSM). Unlike traditional methods that rely on complex transformations such as Fourier descriptors or graph-based techniques, our approach is based on the average boundary distance between two polygons after applying translation and rotation corrections. This method is both easy to implement and computationally light, making it suitable for large-scale applications. The proposed measure demonstrates strong alignment with human perception of shape similarity. However, a notable limitation is that it tends to produce similarity values predominantly within the range of 70% to 100%. This behaviour arises because the measure emphasizes overall shape alignment while overlooking finer local discrepancies. As a result, subtle deviations, such as missing details or minor geometric distortions, may not significantly impact the computed similarity score. Despite this drawback, the method remains a practical and efficient alternative for evaluating shape similarity in large spatial datasets, particularly where computational simplicity and scalability are prioritized. Future works can explore potential refinements to enhance sensitivity to local shape variations while maintaining computational efficiency.