Dense Real-Time Capture of Large Indoor Environments for Immersive Visualization and Telepresence
Jonas Prohaska, Iana Podkosova, Christian Schoenauer, Hannes Kaufmann · 2025
Detailed reconstructions of large indoor environments are needed for many application domains, such as telepresence and as-planned-as-built comparisons for the construction industry. Usually, to obtain the desired accuracy of the reconstruction, a high-end stationary sensor and many hours of post-processing work are needed. Our methodology circumvents this obstacle by creating detailed captures of large indoor environments, including color, in real-time. To accomplish this, we created a dual-sensor setup with a mobile LiDAR and an RGB-D sensor (Azure Kinect), and a dual-layer capture workflow in which a sparse point cloud capture serves as a basis for the second, dense point cloud reconstruction. In the evaluation in which we compared our resulting point cloud captures to the ground truth point cloud created by a high-end stationary LiDAR, our methodology achieved the accuracy below 3 cm.