The context capture is not recognizing the correct height of the bridge deck , firstly I added the tie points on photos and in one AT I got the right height of the bridge deck, but after I added more tie points I only got results with the bridge deck at wrong height.It is important to say that the photos when were taken the shots were not associated with a georreference or geolocalization system, like a GPS, that´s why the use of tie points.The attachments include the settings used to proceed to AT, and images of the results with the bridge deck in the right place and another, after the adition of more tie points, in the wrong place.
What should I do to solve this issue?
Hi Oto,
Thanks for your answer, this setting "orbit around thin structure" really gets better results, now I'm adding fake GCP's to see if works.
The version that I'm using is 23.00.00.1317, I understand that I could get better results if I did another missions to photograph the bridge, but it's not possible, that's why I'm already trying the fake GCP's, anyway in attachment follows two photos taken used to this job as an example.
Which version of Contextcapture are you using?Try setting "orbit around thin structure" block setting so it takes less points in the background.
Overall this exact bridge is not easy to be captured by photogrammetry(repeating pattern issue). You should have set lot of QR tags before so it aligns better.
Also the photos are made too close and AT has issues joining photos in correct order that is why you see multiple components overlapping.
Show a screenshot how was the actual capture plan. You should fly with UAV further away and at angle so the ground is visible(but avoid water) and CC can follow the ground for correct AT. You would need multiple missions at flying 50m, then 30m, 20m and lowering UAV at both ends by 5m steps so it can join together. Same with close zoom you need "zoom by feett"(a.k.a. walk yourself closer) which means that any close photos should connect to photos made further away.
From my humble experience user tie points are not strong enough to correct such issues when AT "thinks" it has found connection but it is actually wrongly connected. Also they need to be spread more apart and if 2-4 user points doesn't already correct the issue then most likely they will not help. You don't need to mark all user tie points in one photo better is to mark at problematic series start and end. Look in Overview page and click on View automatic tie point(s) then you could get better picture where it actually finds these AT points.As last resort you could add manually fake GCPs and then select tie points in 3D view and look from which photos they came and mark GCPs there then these multiple components will merge together by force.
Use fisheye lens or even will have to use mobile LIDAR.