4D BIM and Reality Model-Driven Camera Placement Optimization for Construction Monitoring

Cameras are one of the most valuable sensors for collecting high-quality visual data on construction sites for uses ranging from surveillance to automated information exaction. The dynamic nature of sites means the visual data from cameras can suffer from occlusions and lack of coverage due to progressing works, hindering the performance of automated visual analysis methods. Therefore, manual planning and adjustments by experienced practitioners are required for appropriate camera placement at the site, which is expensive and time-consuming. Past research has simulated cameras and used algorithms with an objective function to optimize installation parameters with planned site models ranging from two-dimensional (2D) to four-dimensional (4D). However, these models lack information from ongoing site conditions, hampering actual camera performance. This study proposes a camera placement framework incorporating 4D-building information model (BIM) and reality models. The framework first identifies the camera placement determinants through expert interviews. Next, the planned BIM and reality models are used to construct the simulation environment, and camera placement parameters are optimized. The proposed framework is implemented and evaluated on a construction site. 25% average camera coverage improvement from the benchmark solution is achieved. This study further contributes to the potential application of automated visual monitoring systems on construction sites.