Mavic 3E and Trimble X7 Alignment

[harrytwomey]

Hello

I'm having huge issues with this workflow, https://agisoft.freshdesk.com/support/solutions/articles/31000168474, 

I have followed the Tutorials and every conceivable combination, but either my scan data which is georeferenced in Trimble Realwork, get moved in crazy location with my drones sparse cloud (laser scan group rest alignment[N/A]), or does not find any matches with the point cloud in the fixed position [R]

Does the E57 you using need to be a coloured point cloud to work?
I tried PTX and PTS with no improvement

My Pointcloud are greyscale intensity and I am only getting 22 match, at most between the scan data and the photos
also, some of the matches are very wrong

When I Import Laser scans there position/rotation (From Origin on import) is in the estimated area? I tried "From Unknown" on import also.

I tried Importing them in the Coordinate system of the photos and local coordinates

I tried all the above multiple times in 2.0.15, 2.0.16 and 2.0.2 with no improvement

I also tried using the layout script but this didn't seem to have any effect.


Attached Picture 1 shows the Low match count,
Attached Picture 2 shows Bad matches and also Valid matches which I would say are incorrect
Attached Picture 3 shows LASER SCAN GROUP being rotated out of position


There seem to be a lot of people here with Mavic 3E, maybe would use laser scan data also, could I DM someone for help

[Dieter]

In my experience, the scan point of view has to be roughly identical to the drone point of view for identical points to be found at all, with colored point clouds working better than grayscale scans. In your example, that's not the case at all. Metashape does not compare the point clouds, but the panorama image, and that should be almost identical to the drone images.

It is irrelevant which model the drone is.

Dieter

[harrytwomey]

Hello, and thank for the response

I see what you mean, I have been working with a limited set of my data for speed, to find something that works

What I actually have is a fully laser scanned building, about 700 scans, but I will limit to just the external that is closest to the facade of the building, which will be around 130 scan

We did a standard topographical flight and built an initial model and used the flight planner to make a path for the Mavik to fly at 7m distance from the building with the intention of making a high-detail model.
We have that processed ok, 3500 photos @ 24GB of imagery, but there are still some areas where the drone could get into, see Picture 1 below(shows bad matches I currently have)

the mavik dataset alone was off 40mm for GCP, whereas my Laserscanner is good to 5mm, so I want to learn how to get the Pictures to align to the Geofrenced Laser Scans, for this and future projects

[Paulo]

My suggestion would be extract foto identifiable GCPS from your laser scan survey and use them to reference the Mavic3 chunk and then merge laser scan and drone chunk....


My worry is why you get 40 mm error on GCPs on your drone Survey done at 7 m distance from building. This would give you around 2 mm GSD and a correct alignment/modelling should give errors around 2 to 4 mm not 40!

[Dieter]

exactly my thoughts.....


Dieter

[harrytwomey]

My worry is why you get 40 mm error on GCPs on your drone Survey done at 7 m distance from building. This would give you around 2 mm GSD and a correct alignment/modelling should give errors around 2 to 4 mm not 40!

Apologies, this was in reference to the standard overhead topographical survey at 100m, which was off 40mm from GCP, vs the Laser scanning of the building which was off 5mm
Oddly, we processed this twice, in DJI Terra vs Metashape and the GCP error was 17mm vs 40mm, but that's an issue for another day.

My suggestion would be extract foto identifiable GCPS from your laser scan survey and use them to reference the Mavic3 chunk and then merge laser scan and drone chunk....

Question: In this case, how would I avoid issues with the two complete point clouds not blending perfectly? The Laser Scans are registered in Trimble Realworks which does not have a scale factor applied, and Metashape, working in my national grid, ITM  espg:2157, with undulation grid to Malin Head datum, would have the scale factor naturally accounted for in the processing. which would lead to a small scale difference in the two clouds. I have tried this before on another project and noticed that flat walls become very rough in the mesh generation because of the thickness of the cloud noise on a flat surface. This is what we refer to as a double skin in the pointcloud, where the two bands of pointcloud data which should be co-planar are offset some bit to cause bad issues.



Im Currently imported all the external Laser scan and all the photo to see if having more options for overlap helps the alignment

Thanks

[Dieter]

We often have this case, there are 2 possibilities:

1) Calculate the photos with local coordinates, then the scan coordinates match the photogrammetric coordinates. This must be agreed with the client.

2) When scanning, set the factor of the transformation directly in the scanner, with our scanner it works. Then the scan is scaled by this factor from the start and fits together with the photos. Unfortunately, this option is not available in Realworks.

Dieter

[harrytwomey]

2) When scanning, set the factor of the transformation directly in the scanner

May I ask what Scanner you use, I see in another topic to mention that you also use the Realworks Software. and generate floorplans. Seem we may be in the same line of work

I have Finished importing all the external scans and in this petical case where we have a generated flight where the photo where taken at close range, the two data sets seem to be aligned pretty well doing nothing else. See pictures of Medium Mesh Generated from Laser and Depth Maps