The utilization of high-resolution 3D point cloud data is becoming more common to a variety of DoD applications and many photogrammetric data sets are collected via small UAVs. Although these data are common, they do not always contain information regarding their geolocation uncertainties. Determination of the geolocation accuracy of any 3D data set typically involves the labor intensive process of an analyst having to identify and extract the coordinates for building corners or other ground targets. Here, we present an automated 3D building corner finder that can be used to determine the global accuracy (i.e. the geolocation offsets) of high-resolution point cloud data. First, building points are identified in both the test and reference point clouds using an automated feature extraction process. Next, building points are regioned into unique buildings and matched between data sets based upon centroid coordinates. Then, an iterative 3D bounding box is passed over each unique building to isolate the building roof corner locations. In addition to the corner finding, our method identifies center ridgelines of rooftops which are useful for estimating rotations about the X or Y axis. Geolocation offsets are calculated on a per-building basis. The output from this process can be directly ingested into the generic point-cloud model (GPM) for the high-resolution data.
Keywords
GEOSPATIAL DATA , METRICS, SYNTHETIC ENVIRONMENT, UAV
Additional Keywords