I have processed a dataset acquired with a DJI Matrice 4E drone, RTK was used during acquisition. I’ve used the coordinates from the EXIF headers for georeferencing, which appear to be identical to those in the .MRK file. I use the Dutch coordinate system (EPSG:7415) as project and output coordinate system. Unfortunately, all output is shifted by approx. 120 meters with regard to reality. The original image coordinates (which are in WGS84 or ETRS89) are in the correct location if transformed within QGIS. This leads me to believe that the transformation from WGS84/ETRS89 (I’ve tried both) to EPSG:7415 is not performed correctly. Have others noticed similar issues with the same or other projected output coordinate systems?
Hi Toby,
what kind of exports have you provided? In which way are the outputs shifted? Have you had set the Project output setting also in the export settings? Also, can you try to use just 28992 Amersfoort / RD New system, as RealityScan doesn’t support geoid heights?
Both the project and output coordinate system were set to the same coordinate system. During the export, I set the Coordinate system to Project output and the Transformation preset to Default.
When using EPSG:7415 or EPSG:25831 (UTM on the ETRS89 ellipsoid) the output is shifted about 110m to the south and 30m to the west. I’ve tried exporting point clouds in LAS format, camera positions as CSV, and an ortho image as TIFF.
I followed your suggestion and used EPSG:28992. This does indeed get rid of the shift, and the resulting heights are very close to the true NAP heights. I find this a bit strange since 28992 is a 2D system.
I guess the alternative would be to first transform the camera coordinates to my desired output coordinate system using the PROJ library and use those inside RealityScan instead of using the original latitude/longitude/ellipsoidal height coordinates.