Scaling - Photo exif GPS

Hello, I have about 2200 drone photos (exif) with 2 components. One of them consists of 2000 photos and the other of 200 photos. Component 1 is absolutely not to scale (RC gives 625000 m while in real life I have about 70 cm). Component 2 is to scale, a priori. These 2 components were made with the same settings.


Merge it with another setting.

And I merge the 2 components and I find myself with a distance of 0.02 m (still for my 70 cm in real life). I don’t understand anything that’s going on.

Thank you for your comeback.

Hi,

can you add here also 1Ds view with the images? Are the images on the right position in the Map view? How do look the alignments in 3D view? Your alignment settings for the components are not good, it should look more like for merging. Mainly Max features per image is quite high.
 

pouvez-vous ajouter ici également une vue 1D avec les images ?

 

 

Les images sont-elles au bon endroit dans la vue Carte ? Pas du tout !

 

Perfect but not at the right scale :

Are you using GCPs in this project? Which coordinate system do you have set in your application settings?

When you just import the images into the project are they showed on the right position (not after alignment)?

Is you set No under Use camera priors for georeferencing in the Alignment settings and then align again is result the same?

If so, can you try clean install as is described here: https://support.capturingreality.com/hc/en-us/articles/4404884840220-Unusual-problems-with-RealityCapture#:~:text=You%20can%20try%20this%20workflow%3A and then repeat the process?

I don’t use GCP but I’ve control point with reel distance.

Coordinate system : espg4326 GPS (WGS 84)

When you just import the images into the project are they showed on the right position (not after alignment)?

Yes

 

Is you set No under Use camera priors for georeferencing in the Alignment settings and then align again is result the same?

No, I’m on YES

It seems that those images near to Africa are causing this. 
Can you go to Map/View/Objects/Camera lasso and select those images? 

Then set Unknow in 1Ds view/Selected inputs/Prior pose

Are those Africa’s images from another camera source?

Are these images of Africa from another camera? NO I have 281 cameras in Africa. I think it's images that did not pick up the GPS... I realign then?

Yes, as I wrote, set unknown for that 281 cameras and then align the dataset again.

OK, I try.

Should I have the correct scale (despite the inaccuracy of the GPS)?

And then I place control points to align the component that is not georeferenced? Will this one fit on my main component scale?

 

After alignement with the 281 camera set unknown, I’ve :

Is this just component from 281 cameras or it is computed in one project? How does it look in 3D view? Do you have created two components in 1Ds view?

When you are using GPS data from the images, the scale will be approximate, not precise. It depends on the precision of used drone.

In your work do you need to have georeferenced data or it is enough to have the right scale? If just scale, then set No in mentioned Use camera priors for georeferencing and align all cameras. Then you can apply a scale to your model.

If there is a good overlap between components it should merge without CPs. But yes, the images without GCP info will be added to georeferenced ones. Also you can set for the images from the georeferenced components as Lock pose for continue.

I'll test it and get back to you. Thanks.
I charge the cameras. I look at the map, and I deactivate the cameras that are not georeferenced. I align the cameras that are georeferenced. (I checked, I'm on the right scale). Then I deactivate my cameras which are OK, I reactivate the cameras which are not georeferenced. I line them up. I end up with several components, including the master which is at the right scale. I create control points to merge the components between them...? Do I align or do I update? I deactivate the camera pior (georeferencement) for my alignment? This is where I'm stuck...

Yes, if you have more components, you need to use control points. Minimally three on minimally three images for each component. After placing the control points it is better to align again.

If you deactivate the camera priors for all cameras you will lose the georeferencing and scale. Just set the wrong positioned cameras as Unknown in Pose settings.

Then place CPs and align again.

Top, thank you very much !