Laser scans component moving position when combined with images

Hi, I’m trying to align a set of images to georeferenced laser scan data, the scan data has been imported as E57 and has had its absolute position locked and ‘lock pose for continue’ enabled.

I’ve added images from both a handheld camera and a drone, set the prior pose to unknown and let it align. I’ve tried aligning with drone images only and hand held images only, both end up with a wrongly positioned combined component.

The problem that I’m having is that the combined model is shifting location on the grid, I’ve verified that this is changing the absolute coordinates too by exporting with my the same coordinate system and comparing to a pointcloud thats not been through RC. The model before combining is in the right location, once merged it has moved position and no longer aligns to the pointcloud correctly.

I went through the CyArk tutorial to see if I was missing anything but theres nothing that I can see. https://www.youtube.com/watch?v=ziHvqz6gTVo

I’m out of ideas for what to try, any ideas on what might be causing it?

Thanks

 

Hi ScanningAHS, when you imported the laser scan data, did you imported it as georeferenced in correct coordinate system? Do you have this system set as Project coordinate system?

Are you using a component workflow or did you just imported images to the project?

Do you have some GCPs in this project?

Thanks for the quick response

“when you imported the laser scan data, did you imported it as georeferenced in correct coordinate system? Do you have this system set as Project coordinate system?”

I have the project coordinate system as local:1 Euclidean and the output coordinate system as my required coordinates.

I’ve tested by meshing the laser scan data without adding any images, when I export that using the same coordinate system I imported it with it lines up with a pointcloud thats not been through RC. Its only once adding images that the position gets shifted slightly out of place.

“Are you using a component workflow or did you just imported images to the project?”

I’ve tried both ways with the image sets to see if that would make any difference but I was getting the same results either way.

“Do you have some GCPs in this project?”

I had control points in the image component when using the component workflow. I did a test just importing the images directly and then aligning, that had no control points. It didn’t seem to make a difference.

 

 

 

 

Can you try to import the laser scans also with set correct coordinate system, not as local and as georeferenced?

By GCPs I though CPs with coordinates, which are placed over laser scan’s dataset and also image’s dataset.

Is it possible to share part of your data for testing?

Hi Ondrej,

I’ve tried aligning with both the project coordinates and the output coordinates as my required coordinate system, this hasn’t fixed the issue.

Just to make sure I’m not missing anything in the settings these are what I’m using:

Laser scan:


Drone images:


I’ve not used any control points as the data all aligns well, the final component looks accurate, however its just in the wrong location on the grid. It shouldn’t make a difference if the scans are locked anyway right?

“use camera priors for georeferencing” is set to yes in the alignment settings

Hi, it seems good for me. Is it the same, if there is set Lock pose for continue as No? Also, if you set Position and orientation and not Locked?

As I wrote, can you share part of your data with us to test?

Hi, do you have an email that I can send the data to? If so I can send over an e57 of the scanned area and drone images

thanks

Yes, you can send it to

One more thing, when you compare Prior pose and Registration pose of the component, is it the same? When you are exporting the mesh/point cloud, do you have set there the correct coordinate system?

Thanks, I’ll get the data sent over to you

Coordinate export should be correct, I’ve verified it by exporting a model generated from the laser scan before adding any images and it was in the correct position.

Had a look at one of the scan positions in the combined component, this is how it looks:

 

Will wait for the data.

According to the attached screen shot it seems that the coordinates are the same. So it seems, there will be some issue with export.

Hello,

I have exactly the same issue here.

Using only local system but with laserscan. The thing I would like to achieve is : be able to align in and out of building based on the geometry from laser scans ; and also separelty the inside ; and the outside on different component while keeping exact same positionning on the grid .

Everything I have tried did not succeed (i kind of tried a lot) and there is always some shift of the differents components on the grid.

Any idea ?

Hi NordHeritage,
regarding to the local system, all of the scans are in the same local system or inner and outer scans have their own local system?
Are the inner and outside scans registered/connected into one rig?
Are you using control points or ground control points? If so, are they measured (do you know their exact coordinates)?
Do you want to combine the data with images or do you want to use just laser scans?
What kind of laser scanner are you using? Is the registration provided before import to RealityCapture? What are your import settings? Do you have set the coordinate systems correctly inside RealityCapture?

Hi NordHeritage do you have solved the problem?

Hi,

unfortunately my issue has not been resolved.

We are using RTC 360, registration made on Cyclone. We then import the e57 as exact. We simply want to register drone (M3E) and ground images (A7R) on the first registration made by the laserscan alone. We are using components workflow and locking the e57 one as exact in the relative pose setting and lock pose for continue.

The issue is that the flatness of the “master” component change after merging all of them. We would like it to stay as flat as the reference i.e. the e57 component’ flateness.

Alignement settings are the default settings.

(local euclydean coordonate system ; no controls points nor GDP as we trust laserscan for geometry)

Hi
About the registration in Cyclone, are the scans merged into one component or do you have separated outside and inside component?
If it is separated, are your scans in their own coordinate systems or in the same local coordinate system?
If you want to keep the laser scan’s position in RealityCapture, you need to import it as Exact and georeferenced in local coordinate system (if you are not using some global system).
And if you want to keep the laser scan’s system, then you need to set Unknown for a prior pose of the images.
As you are using component workflow, do you have only laser scan and images components, or are you separate it more?
After proper import it is not necessary to lock or set lock pose for continue (this is just locking the relative pose between cameras, not the absolute one) for laser scans.
Can you show the flatness change here?
Are you able to merge scans and images without control points?

Hi,

Thanks for your answer ;

From Cylcone it is an only one big component (inside outside).

The scans are in local as well so I tried what you mentioned and applied import with georeferenced on and local coordinates.

There is still a change in the “flatness”.

Regarding the workflow : More precisely, after having imported the scans, we run a first alignement ; and then directly add images on the same reality capture opened window and re run alignement.

Sometime we add all the images (little projets) ; sometime (complex project) we need to run several time this process (import scan ; add some of the images project ; rerun alignement ; export registration of the area) and then do the same thing with all the others parts of the project.

The original LSP are always the same

Sometime the levelness is ok ; sometime it shifts a little .

(We are trying to use as few control points as we can - so we use them almost never).

Hi, can you send a print screen of this flatness?
Also, can you check the prior and registration pose for the LSPs, if those were changed?