Im looking to create a new video on how to make the best possible ortho image from UAV photogrammetry, In terms of efficiency and quality.
?ÿ
I was hoping to get some input here from this fine community if anyone has any tips and tricks I can add to the list of parameters I am going to test?
?ÿ
I have an idea on what the answer is to most of the following questions but was hoping to get opinions of others, and then do some testing to hopefully get a somewhat definitive answer.
?ÿ
How much motion blur can you have before it affects quality? 2x GSD, 4x GSD?
At what point does shadowing become too significant an issue?
Does more overlap at a higher altitude help with stitching elevated objects such as powerlines and tall buildings?
Does more overlap improve quality? What is the minimum overlap required for a typical (non homogenous) area of interest?
Do obliques images help?
Does creating your DEM from depth makes make a difference?
Are there any processing settings that make a significant difference (blending mode, refine seamlines, hole filling, ghost filtering, ect)?
Can you create an acceptable ortho without any Nadir images?
What effect does changing your f-stop have? What is the best f-stop to use? And in what situation?
How much effect does changing your iso have? Can I get away with iso 200, 400 in low light conditions without significant quality loss?
?ÿ
?ÿ
If you can think of any other questions you'd like an answer to or to have some testing done, please add them here.
Interesting..
We used Orthos mainly as one of the internal sources to digitise & such, and somewhat infrequently as deliverables.
We were told several times to generate our orthos using DTM (ie bare ground), but in our own experience, Orthos generated from DSM tends to be a little better esp. for structures such as buildings and trees..
We now fly both Nadir & Oblique at 60 or 70 degree when possible, build DSM from that and generate the ortho using the DSM.
It would be interesting to get a verdict or what is the actual story on this.
.
P.S We use P4RTK and Metashape, and located about 3 degree above the Equator. Let me know if you require some data at certain parameters etc so we could chip in in providing the dataset and such..
Regards
At what point does shadowing become too significant an issue?
Shadows have been a huge problem for us. Basically anywhere there’s a shadow is garbage data (in my experience). We can do our best to find reasonable guestimates of where ground “ought” to be in TBC, but just saying that out loud makes me cringe and any of our shaded areas really just need to be surveyed.
We’re using a P4RTK with the standard sensor and the flight software that came with the controller, generally flying about 75% of the recommended max flight speed, and we have the shutter speed(?) set for 1000 when sunny, 800 when cloudy. We process in Pix4D.
I’ve been dreaming up some ideas about flying additional lower/slower flights with ISO settings optimized for shade, etc, but there are plenty of reasons that is challenging (including my boss who would want to know why it took so long - trying to get better data and save on office time, duh, bruh).
Can you create an acceptable ortho without any Nadir images?
This is a yes for us. We found flying the standard 3Dgrid mission is quite sufficient to generate an acceptable Orthos, no significant difference from the one one generated from Nadir that we've observed. Soft photogrammetry has come a long way..
We have experimented with lots of different settings until we got a reasonable SOP for orthos. I have a background in photography from the analog/film days so always had an instinctive grasp of fstops and shutter speed.
Ideally we want to fly on bright overcast days but when we fly on sunny days we try to expose for the shadows in order to get some detail there. We start with not flying faster than 7mph and set shutter speed at 1/1200 or 1/1600. then we adjust fstop to get as closed down as much as possible so 5.6-7 if we can. Last we set the iso to get a +1 on the ev scale. So the iso can be anywhere from 100-1000 to get the right setting. Iso is only an issue in low light situations where you introduce noise.
We use an Autel drone which has a rolling shutter so flight speed and shutter speed are the critical variables. fstop and iso are less important. In fact as the mission starts, I watch the evscale and just adjust the iso on the fly to keep the +1ev.
The final ortho uses a process to create a seamless gray scale of the stitched images. Yes, we get blown out highlights on concrete and roof tops but we get detail in the shadows by the building which is more important.
We also now use red paint to mark our gcps/trav points, as you previously suggested. Shows up much better than pink or orange.
