Abstract

As-built Building Information Models (BIMs) can significantly enhance the efficiency of facility management and operations. However, they are seldom used in the architecture, engineering, construction, and facility management (AEC/FM) industries due to the difficulty of collecting the spatial information required to create such BIMs. Typically, this spatial information is collected through manual field surveys, which are time-consuming and error-prone. Therefore, image-based sensing techniques have been proposed and tested for the three-dimensional (3D) modeling of spatial information using images captured by handheld cameras. Moreover, with recent technological advancements, Unmanned Aerial Systems (UAS) equipped with high-definition cameras have become a powerful tool for remote image acquisition.
So far, it is still not clear whether images remotely captured by a UAS can be used for the 3D modeling of indoor building environments. The main objectives of this research were: (a) to test the feasibility of using a UAS for the 3D modeling indoor building environments, and (b) to evaluate the effectiveness of using the UAS for 3D modeling in terms of accuracy, density, time, and cost. In order to achieve these objectives, a UAS is first deployed to collect several overlapping images of an indoor building environment. Next, a Structure from Motion (SfM) algorithm is applied to generate a 3D point cloud that represents the spatial information of the indoor building environment. Later, the accuracy and density of this 3D point cloud were evaluated by comparing it with the ground truth point cloud generated from a laser scanner. The time and cost needed for modeling using the UAS were also recorded and compared with the ones from the laser scanner. This framework was applied in two case studies: (1) the 12th floor of the Hall building, and (2) the 2nd floor of the EV building, at Concordia University. The results of these studies showed that 80% of the 3D points had a deviation of up to 8.8 cm from the ground truth. Approximately 20% of those points had a deviation of 0.81 cm or less. Overall, these studies demonstrate that UAS-based modeling is able to model the spatial information of indoor building environments with dense point clouds and with acceptable accuracy. Furthermore, the studies also indicate that the 3D modeling of spatial information with a UAS can be a time- and cost-effective approach.