USE OF KINECT AZURE FOR BIM RECONSTRUCTION: ESTABLISHMENT OF AN ACQUISITION PROTOCOL, SEGMENTATION AND 3D MODELING

With the popularization of RGB-D cameras, access to the third dimension is now possible at low cost. However, these systems have a lower accuracy compared to other technologies such as terrestrial laser scanners (TLS) or mobile laser scanners (MLS). RGB-D cameras have proved their potential for 3D indoor mapping, particularly for Building Information Models reconstruction (Li et al., 2020). This paper aims to investigate the acquisition protocol and propose a method for BIM reconstruction by using an RGB-D camera (Kinect Azure). First, an acquisition protocol is established with the aim of improving the quality of 3D reconstruction of indoor scenes. Different scene cases are considered, namely a single wall, a corridor, a room (with different levels of detail) and two adjacent rooms. After having extracted the best acquisition scenarios for each case of the studied scenes, an image processing is performed for the most complex scenes. Then the 3D reconstruction is performed and the resulting point clouds are subsampled and cleaned. Next, an evaluation of the geometric quality of the 3D reconstruction is performed, by making a comparison between the point clouds from the acquisition protocol (room and corridor) and the reference point clouds from an MLS. The results of this comparison shows that the differences between the two point clouds have an absolute average deviation that doesn’t exceed 4.8mm, which proves that the proposed method has reached competitive accuracy. Finally, segmentation and 3D modeling of the studied scenes are proceeded to extract the BIM objects