Hi All!
Currently processing a model with 3 GCP's. I have imported the co-ordinates and selected the relevant co-ordinate reference system. When I pick the points in the editor for 2-3 photos, they appear in red. The GCP's successfully import after I pick my points. However, after running AT, the model appears to be floating along the Z axis. Is there something I am missing?
Do you have good distribution of GCPs? How the photos and their positions were acquired?
Also try to run first aerotriangulation as accurate metadata.
Hello Otto,
For the sample area under processing, I only used 3 GCP's and they were staggered across the site. The photos and positions were captured with Drones.
For the entire stretch to be surveyed we decided on 10 GCP's well spread around the corners and through the alignment of a highway corridor under construction (as depicted in the image below). The entire stretch is 15kms. Based on your experience, how many GCP's would you suggest we install to get an accuracy of 2-3cm and for ContextCapture to easily process as well?
I had not experience with road surveying but wisdom is that more GCPs are better. Be aware that if first aerotriangulation is not good it can cause troubles as we found in this case - https://communities.bentley.com/products/3d_imaging_and_point_cloud_software/f/contextcapture-descartes-pointools-forum/154013/issues-with-a-produccion-after-reconstruction
From my experience most issues with GCP is when photos are blurred(out of focus or speed too high) or rolling shutter distortion. CC can't handle blur so the model will not be accurate and adding more GCPs won't help only make it worse even breaking model in seams. So go trough delete all blurry photos and also photos which are taken while drone is rotating. What drone and camera is used? Fly height and speed? Can you attach Quality report from Contextcapture?
Couple of things .. 2 to 3cm accuracy from drone for this sort of job is optimistic .. the basic rule-of-thumb is Accuracy = 3 x GSD ... so you are shooting sub-cm pixels - for 15km of road .. probably with cheap(er) cameras/lenses .. that is keen.
Secondly - you can only be as good as your control .. how are you getting the values for the control? RTK GPS? Static GPS for 6 hours+? Total station? If RTK - you have more than 2 to 3cm in the control alone .. 2 to 3cm in X/Y .. up to 4cm in Z.
Moving on - try running the AT first up on the imagery without the control - this assumes there is at least some level of GPS on the drone(s). You should at least get an AT that resolves ..
Now add your control to the block .. you should get red circles in the 3D view where the control should be .. are they at least in roughly the right place? If not - check eastings versus northing - numbers in the correct columns? Correct coordinate systems? Hover over the red circle you should get the control mark ID - the one your are expecting?
If this works .. try nominating 4 control marks in 4 photos each and run it .. this should plonk it roughly in the right place .. makes the hints easier to pick in the imagery.
Pick more on the next AT etc ..
We use a DJI Inspire 2 with a Zenmuse X4S camera for data acquisition, flying at 70m height and speed of 7m/s. We haven't had any blurred photos at the height or speed we fly. We understand that for such a large area, a fixed wing drone is more feasible which we are looking into for future use. But for our current POC, we've gotta make do with what we've got. So the question about how many GCP's for a certain area is a real issue considering GCP's are pretty expensive in this part of the world.