How can markers be off target on orthomosaic if errors values are good ?!

[jkova96]

I want to know if can someone explain me this,
 
Like said in the tittle of this forum, I want to know how can markers be off target - I mean on photogrammetry targets which is used for GCP (Check points) measurements, if errors are good?
I really don't know how can this happen? In my case I figure out this after optimizing photos and comparing how markers lays on orthomosaic.
Check out for pictures shown below - 1.st one - markers and photogrammetry target over 5 cm orthomosaic. Size of photogrammetry targets is approx. 40 cm by 40 cm.
2.nd picture shows my values on reference tab.

I did compared results via Pix4D Mapper and on that software seems everything OK.

Have a nice day !

[DayGeckoArt]

Is it possible the GPS coordinates are less accurate than you think? Also did you try loading the orthomosaic and GCPs into GIS software?

[jkova96]

Hi DayGeckoArt,

They should be accurate.
It's used GPS RTK method. Also, coordinates from drones are recalculated by using GPS STATIC method and combing these two it should be definitely as good as you thing.
Average accuracy should be within 5 cm.

After receiving email message from Alexey - I think what might be the problem. I imported existing orthomosaic from another project which have higher estimated errors - after optimizing cameras.
In fact - even for that project no matter cause errors are high - points falls on exact place where it should falls - in the center of photogrammetry markers. I can share with you images how they falls via this link: https://mega.nz/file/09JSiBrZ#ASC4lQAaUzLi1hf3h2U6AxB6fX9yYIrtZiXLxLQtTx8

For this project errors are quite smaller comparing to that one.
In addition to that check points falls way over photogrammetry targets used for GCPs in the field.

I think that this fact - imported orthomosaic from project with higher errors after camera optimization might be the reason why now I have shift on check points from where it should be. I can never be sure before I see actually that overlap.

I'll note you when I get good, or in my case "obvious", results.

Thanks !

J.K.

[tazzo]

My 2 cents:
- Your GCP are too small and pixeled to be accurate
- You need to manually center all blue flags and also double check the green ones, GCP are too small to be automatically centered in a good way
- Absolute GPS error (even with RTK) is likely much more than 5cm

[jkova96]

Hi tazzo,

It's not to small targets. Around 50 cm. I don't know, it depends how high you fly and how much accurate you need to be.

It's pixeled cause I exported less quality orthomosaic from another project and imported to this one - in which I have problems, or "problems" because I made it by myself cause that orthomosaic came from 1.st project after optimizing cameras and exported into 2.nd - and it was 5 cm = 1 px.

Originally you can set, maybe 1.5 cm = 1 px...but when I'm having smaller numbers then much heavier file I got. So, I always try to set accuracy lowest possible for (re)importing...cause I want much faster file import.

By the way - what did you mean when you wrote this: "You need to manually center all blue flags and also double check the green ones, GCP are too small to be automatically centered in a good way". Cause, when you look at the orthomosaic, in orto view, you can't actually move any marker anymore. Something similar I saw from Alexey, but can you describe more? I can, i think so, manually move markers (blues ones) when I'm in photo view mode, not in the "ortho" view mode (I mean there are three modes - ortho, model and photo). What about white/grey markers?

My second question to you, in addition to above, would be, cause - I see that you understand, or i think that you understand procedure with GCPs:  When doing that after "optimizing cameras", cause all of this happen after optimizing cameras, should I optimize cameras again (one more time)?

I optimize cameras once in the project after I place all GCPs on pictured with, usually, green markers? So after that, should I check for the - only blue ones and then 'reoptimize' cameras once again?

Hope you understand everything I write.
Thanks !

Bye for now

J.K.

[tazzo]

If you fly so high you need bigger target. I know it's not super practical, but 1 meter targets may help you in this case. Also be sure that they have a lot of contrast.

Let's say you want things accurate.. you put in your scanning area 50 targets. Half (25) are used in the alignment, half (25) are to control the accuracy. After doing "Tools/Markers/Detect Marker" always double check EVERY marker, because the software can be wrong in the detection and may need some adjustments. You don't have a QR code so be sure that marker's name match positions you import.

When you do the alignment, camera are optimized, after you can remove some points you are not interested and do the "optimizing camera" again.

Note: Blue flags means that position is ambiguos and MUST be manually checked, Green flags means software confidence is higher, but I suggest you to double check them too.