Geometric comparison and quality evaluation of 3D models of indoor environments

The increasing demand for automated, cost-effective and time-efficient indoor modelling methods leads to a need for performance evaluation of these methods by assessing the quality of the reconstructed models. In this paper we introduce a method for geometric comparison of a 3D indoor model with a reference, which is useful not only for evaluating the geometric quality of the model, but also for change detection and temporal analysis of the building. The method provides suitable criteria for the quantitative evaluation of the geometric quality in terms of completeness, correctness, and accuracy. Experimental evaluation on a synthetic dataset and the ISPRS benchmark dataset shows the potential of the proposed method for quantitative evaluation and localization of geometric errors in 3D models of indoor environments.