I really need some help to figure out why checkpoints error so big.
I used 17 GCP’s, 4 of it check points. GCP were placed with GPS RTK with accuracy 2-3cm.
If you look at report X error about 10cm, Z error about 20cm.
What I am doing wrong? Any ideas?
I also used structural mode. The result a bit better but still 10cm horizontally and 10cm verticaly.
It looks like you did everything right, but next time you fly a location like this I would recommend disabling “enhanced 3D” when capturing the data. This mode is designed for buildings and structures, and as a result will lower your accuracy in terrain type scenes.
Your GSD is 3.5cm, and so as a rule of thumb the best possible accuracy you could expect is about 7cm (2 times GSD). The digital terrain model generated was 12cm GSD (always a bit lower than the RGB image) so again, you are close to the limit of what is possible, unless you fly at a much lower altitude.
This is what I experience when trying to use GCP’s and obliques and structural mode. Not nearly as bad, but it is never as good as nadir-only. As James said, your GSD is not going to allow you to get better than 5-7cm either way so it just compounded. In my opinion you are going to get good ground data or good structure data. You can run a model with obliques in terrain mode which is better, but I have never reliably gotten both within our tolerances with obliques and structurals.
Thank you James.
What are parameters can be to get better result? Flight altitude 60m with 85%-80% overlapping?
Yes, flying at a lower altitude increases GSD, but you will have to make your own decision on what is acceptable and efficient. I would think 70m with 80% front and 75% side overlaps would be just fine on a project like this. Has someone staked out and confirmed the checkpoints? Verified that the road height was correct? You can’t always assume that ground survey data is accurate when something like this happens.
I have also found that because the point cloud is so dense that you can move around the checkpoint area and the surface can easily change a couple of cm so it is very dependent on how/where the point was shot versus how the photogrammetry interpreted the ground at that point. To me it looks like everything is fairly consistently low so I would look somewhere else for fault.
I will try it on my next flight . For sure will stake out created surface from flight.