Benchmark Dataset for Multi-Source Building Footprint Association (Cologne) - MatchGT

Abstract:This dataset serves as a benchmark for multi-source building footprint data association (matching). It is derived from the "Cologne" experimental area used in the research paper: "MatchGT: A Candidate-Pair Competition Graph Framework for Multi-Source Building Footprint Data Association". The dataset presents a high-overlap, urban matching scenario with minimal non-linear displacement, suitable for evaluating global optimization and graph-based matching algorithms.Data Content:The dataset contains two shapefiles representing building footprints from two distinct sources:Source Dataset (ALKIS): Official cadastral data from the German Official Real Estate Cadastre Information System.Target Dataset (OSM): Crowdsourced building data from OpenStreetMap.Ground Truth: Manually annotated matching relationships identifying correspondences between Source and Target (including 1:1, 1:n, m:1, m:n, 1:0, and 0:1 relationships).Study Area:Cologne, Germany.Number of Source Buildings: 313Number of Target Buildings: 354Number of Matching Groups: 298File Structure:demo_1.shp: Source building polygons.demo_2.shp: Target building polygons.Match_Label.csv (or included in attributes): The reference matching labels.Usage Notes:This dataset is intended to ensure the reproducibility of the MatchGT framework and to serve as a testbed for future research in geospatial data fusion.Citation:If you use this dataset in your research, please cite the associated paper:Qiu, Y., Wu, F., Qian, H., Yin, J., Liu, C., & Gong, X. (2026). MatchGT: A Candidate-Pair Competition Graph Framework for Multi-Source Building Footprint Data Association. Information Fusion. (Under Review).