Author Topic: Use markers to correct shifted point cloud?  (Read 1733 times)


  • Newbie
  • *
  • Posts: 7
    • View Profile
Use markers to correct shifted point cloud?
« on: July 01, 2019, 01:15:13 PM »
Dear community,

I want to create an georeferenced point cloud for vegetation analysis. The point cloud will be used to get information about the absolute height of vegetation.  I imported six oblique aerial images which overlap with my area of interest. The reference system of these images is EPSG 28992. I have set this one in the Reference pane. Furthermore I imported camera positions of each image (x, y, z, roll, pitch and yaw).

I set the camera type to 'Frame', focal length to 100.5 camera size to 0.006 mm x 0.006 mm.

Then I alligned my six aerial images. After that I optimized my camera. I got a RMS projection error of 0.06.

I continued with building a dense cloud (High accuracy, mild depth filtering, calculate point colors)
I exported the results as a las file.

I opened the las file in ArcGIS Pro and visualised it on top of aerial imagery background map referenced with coordinate system EPSG 28992 (RD New).

 It seems that the point cloud is not exactly I expected it to be. Locations with buildings for example, where I would expect higher elevation points were not represented well. The high elevation points were shifted for example approximately 5 m to the left and 3 m southward.  There seems to go something wrong..

Hence, I am thinking of useful solutions to get the point cloud in the right place:

- I am thinking of creating markers on my aerial images including x, y and z coordinates. At this moment I did not use markers in my Agisoft workflow. Maybe the referencing of my images will turn out to be better when using markers.

Do you have any suggestions about my workflow and how I can get put the point cloud in the correct place?
Do you for example have ideas how I can generate manually markers with x,y and z coordinates in QGIS or ArcGIS?

Thanks in advance for you help and reply!

This image shows the shift in north-west direction, the building is not represented completely. The grid shown is a 5m raster, dynamic scaling.

« Last Edit: July 03, 2019, 11:17:55 AM by Lambertus »