Hi everyone,
I was interested to know if RC supports the importing of geometry of a small section of a film set (in my case a small section of LiDAR as an .obj file) so that it can help the alignment of photos and reconstruction of a larger section.
I was able to obtain a small section of the set (not the point cloud, only resulting mesh) and but luckily a good coverage of additional photos was taken (with a hand held DSLR), While the reconstruction of RC has done a good job of recreating the set, would it be possible to use existing geometry to help the alignment process as well as scale?
I couldn’t find a mention of this topic anywhere, as when we talk LiDAR most people are using the the point cloud for reconstruction use in RC, which sadly I do not have the data for.
Cheers!
Hi patrick,
you need ordered pointclouds, that means you need the points as measured from each position of a stationary laserscanner. Anything else won’t work with RC.
Thanks for the information Götz – I will see if I can get that from the vendor (though highly unlikely at this point)
Once again, thank you
I found that if you import your model in your 3d software and snap locators ( dummy etc depending on the name in your 3d software) to your model, you can use the coordinates of ythose points in RC as ground control and it will match perfectly. (you just need to make sure you work in the same world space as your obj.
Hey peregrine,
That’s a nice workaround!
Is it precise, as in the errors are 0 or is it “only” an approximation?
Hey Gotz,
Unfortunately, it seems like the current version is bugged (it was working in a previous version).
When I try to switch the tie points to ground in the lidar component, the points shift, even when I try to force the coordinates of the points.
Then when I re-align, I get error messages on the CPs.
That s a shame, it used to work nicely.
Noooooo. This sounded like a such a nice solution to importing unstructured point cloud data! Hopefully we can get it working again soon.
While that is a nice work around, I’m sure I’m not the first person (or last) that has run into this situation. Users are not always given the ordered pointclouds as vendors tend to loose, charge extra etc etc for that data.
If RC could have the ability to import an model (obj, fbx, etc) and process the vertices of that mesh to an ordered pointcloud, that would be a super useful tool. Then based on that pointcloud not only would your alignment be far more accurate you could also get the scene scale correct, straight out of RC which in my line of work is paramount.
Wait a second, guys. Importing a mesh or a pointcloud is absolutely possible. RC will import it according to its orientation, whatever that may be. I think there is some confusion going on between importing a laserscan and importing a model - I only spotted it now while re-reading. The workaround is meant for meshes (“model”), something that the TO is not interested in.
For laserscans to be used for orientation of the cameras (images), ordered pointclouds are the only way. Some people have managed it with unordered ones, but in my opinion that was only possible for very specific cases, where the geometry was not very complex and the center point was so that RC could use it as the position of the scanner and so make use of it. With very complex geometry, that won’t work since RC will still use the center as the scanner position and so it will be occluding most of the geometry and an alignment should be impossible.
