An Approach for Occlusion Detection in Construction Site Point Cloud Data

Data collected using laser scanners on construction sites often include regions in 3D space that cannot be observed beyond occlusions, which are objects in the line of sight of the scanner. These occlusions may exist even if scans are planned using a scan-planning algorithm. The issue of occlusion can prevent accurate modeling of objects in a scan, requiring potentially costly decisions to revisit the site for additional scans. Computational support is needed to help quickly decide whether obtained data is adequate, or if additional data collection is needed to meet scanning objectives. This paper describes an approach to rapidly interpret point cloud data obtained from construction sites. This approach can help determine whether to collect more data, to use modeling techniques to identify features or objects in the existing data, or to continue without data in occluded spaces. The paper demonstrates initial experimental results obtained by applying this approach to simulated and actual point cloud data.