Automated image-based reconstruction of building interiors – a case study

Abstract

In 3D reconstruction of indoor environments, automated methods offer possibilities for e.g. brokering, planning and decoration businesses. The application potential of these automated methods is, however, tied to the accuracy of these methods. In this paper, we perform a technical case study analysis on a state-of-the-art image-based method. For accuracy, we find that 60-70% of points in the reconstructed 3D point cloud are within 5 cm error range. Image-based 3D reconstruction thus offers potential for those various indoor-related applications that are satisfied with this level of accuracy. We also discuss other factors affecting to the applicability and robustness of this method.