Hi,
We are using a camera rig to capture scans and are therefore looking to utilise the use of XMP for alignment. However, when using the XMP data it results in a poor-quality resulting mesh. Also, if I use the detect marker tool (markers were used in the generation of the XMP data), the projection errors of the control points will be way over the limit, which I’m assuming is leading to the poor mesh quality. Any idea why this is happening or how to overcome this?
Thanks,
NB, I followed this tutorial to generate the XMP data:
Hi r42b42,
how many images do you have in the rig? What kind or rig is that? Are all your settings exactly the same as in tutorial?
When you try mesh your project without XMPs how does the model look?
How precise were your GCPs created/measured?
The rig uses 112 images, 7 cameras on 16 poles surrounding the centre. I’ve used the same settings as the tutorial. Please see the attached images for a comparison between the mesh without and with using XMPs. The GCPs are a known dimension ±3mm.
±3mm is not so precise in this case and it can cause this kind of issue.
For second model are you using a new project? Are the images the same or those are other images?
Is the rig stable between captures?
What are your alignment settings?
Have you tried to use also other XMP’s export options?
This is a separate project to which the XMP data was taken from. The source images are the same, just with and without the use of XMP data. The rig wasn’t touched in between captures.
Alignment settings used are:
Max features per mpx: 10,000
Max features per image: 80,000
Image overlap: Medium
Image downscale factor: 1
Max feature projection error: 2.0
What other XMP export options would be worth trying?
Not exact, as you have ±3mm precision, which could be 6 mm on some places, which is a visible movement. So, export as draft in this case.
But also, it would be good to see your data.
Thanks, using export as draft produces a much better mesh surface. Out of curiosity what is the difference in exporting as draft compared to exporting as locked?
As draft are used the camera position only as suggestions and they are changed during alignment. Exact uses exact coordinates in the alignment process, but as it is not so precise for you, you will get those misalignments.
