This paper presents a new method and experimental software for automated construction progress monitoring. The algorithm is based on an as-planned Building Information Model (BIM) which is compared with the 3D point clouds from a static Terrestrial Laser Scanner (TLS). The point clouds are measured on-site during the construction process. Due to the to the efficient combination of measurement and model, the construction work can be carried out more quickly and with fewer defects.

A vital prerequisite for construction progress monitoring is the co-registration of the acquired data. If this step in the processing chain is biased, all computed deviations among BIM and captured data will consequently be erroneous - false conclusions are inevitable.

In contrast to other solutions our approach is plane-based for both, the point cloud and the building model. The reason why planes are used instead of points can be justified by the large amount of data is reduced to plane parameters, which drastically reduces the required data volume. The small plane sections used for this purpose are called patches. The presented method considers the error budget of the scanner, its calibration and registration by variance propagation utilizing stochastic tests. The building model is reduced to component surfaces called faces. Construction progress is derived through a comprehensive comparison of patches to faces. The position of the laser scanner in the scene is also used, which gives additional information on the visibility and orientation of the building components. The basis of the method are detected patches in the recorded point clouds as well as planned building faces that need to be transformed into an identical reference coordinate system. Therefore, a new method for the co-registration of point clouds into a BIM-coordinate system was developed, whereby the co-registration directly extracts planes from the building model. The new methods are validated within an actual construction project where three successive construction phases were monitored.