I had 2 project files, each containing a laser scan (done with ipad pro lidar with Dot3D) that were imported as exact registrations.
They are of a mine level/tunnel so there is no georeferencing of the scans. I have aligned both separately with images I took. Then I have exported one of them as an alignment component, and imported it into the other project. And added control points and then tried to merge the components. They finish aligning and the scan parts are basically sat on top of each other in the middle of the ground plane and ignoring the control points, whatever weight I give them.
I had previously tried doing both scans in 1 project file (instead of doing seperate and then importing), the result had been the same as this.
It would seem I need the imported component, atleast the scan
parts, to not be locked in position. The imported scan needs to stay together as a whole and it doesnt need to change anything other than the position and orientation/angle, and not be locked in place on the ground plane. Anyway I can do this?
Hello @DeycallmeKoles
This is quite unusual user case for such device.
How many LSPs are created after the scan is imported into RealityCapture?
As you have two scans and they are not georeferenced, they are using their owns coordinate systems. In this case you need to import those scans as Unregistered and merge them using control points.
Also, as you have aligned images, you ca try to use this workflow (or use its principles to get the scans into one coordinate system): https://www.youtube.com/watch?v=60wLmup4y00
It is indeed. But i find photos alone of the mines are really problematic to align, and end up with the wrong scale and proportions, and with lots of gaps. The scans fill those gaps but also give a foundation structure that I can align the pictures with.
The main component has about 200 lsp’s from the scan, and the imported one has like 700 I think.
They are aligned with pictures yes, which I hoped would help, with merging. But they are obviously locked. I’ll have a look at that video on my lunch brea, thanks!
I will try importing them as unregistered. But is there no way to do it retrospectively? Just, these arent the only 2. I have a model of this mining valley which used photos, and then I have 2 completed models of differnet mines each made up of a scan and photos, which I was planning to merge. This is the third one I was making. And I would like to not have to start again.
Alternatively, I guess I could merge each scan and photos to create individual models and export that, and align in cloudcompare or something, then export and reimport in RealityCapture. Just have no idea what settings I should use for that, and I assume I’d reimport it as a model rarher than a laser scan.
Unfortunately, when you change the import settings, then new LSPs are created, so you need to start over.
Regarding your alternative way, that is also the option. In that case you can use the general settings. And for that you can also find some tips in the tutorial above.
Yes I saw the tutorial above, that seems like a good way of doing it. I had tried that a while ago with the lidar scans themselves but didnt have much luck in getting it to line up nicely, but i think the exported models from RC aligned with photos will be easier as they have more detail and identifiable features to align them correctly.
Can pictures be aligned with imported models? Not that I need to do that, but curious if you can. And also I noticed last night that individual lsp’s have settings to change relative position (currently set to exact), but it can be changed to draft or unknown. There is also absolute position and locking group in which you can type a name. What do these do? I assume this wouldn’t help me.
It is not possible to align the images with the imported model.
You can only texture the imported model, but that needs to be located in the same coordinate system as the model created from the images (as it is mentioned in that tutorial).
Regarding the camera priors settings, please check our Help: RealityCapture Help
Okay, I thought that was the case.
And yes that makes sence.
I don’t entirely follow about the camera priors.
When I import a laser scan the registration options are:
“•Exact Imported poses will be preserved; in this case the imported model defines the scene coordinate system.
•Draft Scanning stations have been registered but the registration needs to be fine-tuned.
•Unregistered When the stations have not been registered, the registration engine will calculate poses of all scans.”
Would changing the relative pose of each lsp to draft or unknown, and then aligning, not undo initially selecting the exact registration? Sorry I just don’t understand the difference.
The imported LSPs behave as images/cameras in RealityCapture. Therefore the setting are the same here.
So yes, if you change these after import, the initial setting won’t be applied.
I attempted to import the laser scans as draft and as unknown registrations, I was able to get most of the scan LSP’s to align, but not all of them, and couldn’t get it to align with photos at all.
I actually forgot to try aligning two laser scans, so will try that later. But the fact it won’t align with photos will be an issue.
I’m wondering if importing them both as exact, and aligning, as seperate projects, but then the scan I want to merge with the other, if I set all the LSP’s of that one to draft or unknown, could I use locked pose group to lock them together or would that lock the coordinate position too? Surely there must be some way to lock them together but not to the coordinates position.
You will need to use control points to align images and scans in such case.
You can try to import the scans as exact, but set the Georeferenced to No. Then you can use some control points for merging those (it should work). I suppose it will only name that group of inputs with the same name. To lock the relative poses in the alignment you should use Lock pose for continue as Yes:
