GEO-REGISTERING CONSECUTIVE DATASETS BY MEANS OF A REFERENCE DATASET, ELIMINATING GROUND CONTROL POINT INDICATION
- Dept. of Civil Engineering, TC Construction - Geomatics, KU Leuven - Faculty of Engineering Technology, Ghent, Belgium
Keywords: Geo-registration, Co-registration, Construction, Ground Control Points, Remote Sensing
Abstract. The architecture, engineering and construction (AEC) industry’s interest in more advanced ways of regular monitoring of construction site activities and the achieved building progress has been rising recently. This requires frequent recordings of the area. This is only feasible if the profound observations only require limited time, both for the actual capturing on-site as well as processing of the recorded data. Moreover, for monitoring purposes, it is vital that all datasets use a single, unique reference system. This allows for an easy comparison of various observations to determine both building progress as well as possible construction deviations or errors.
In this work, a framework is proposed that facilitates a faster and more efficient way of co-registering or geo-registering consecutive datasets. It comprises three major stages, starting with the capturing of the surroundings of the construction site. By thoroughly adding numerous ground control points (GCPs) in a second phase, the processed result of this input data can be considered as a reference dataset. In a third stage, this known component is used as additional input for the processing of subsequently captured datasets. Using overlapping areas, the new observations can be immediately transferred to the correct reference system. This eliminates the indication of GCPs in subsequent datasets, which is known to be time-consuming and error-prone.
Although in this work the focus of the proposed framework lies on a photogrammetric recording approach, it also is applicable for laser scanning. Its potential is showcased on a real-world apartment construction site in Ghent, Belgium. In the test case, the presented approach is shown to be efficient, with comparable accuracies as other current methods, however, requiring less time and effort.