Duplication effect on textures

Hello,

I am currently working on a project where orthoimages must be really accurate. However, I have been experiencing some strange duplication effects on textures on several places of the model. See the image attached.

Is it related to misalignment issues, as I also struggled to align scans and drone images (a lot of double walls and needs of control points)? Is it solvable? Maybe by changing the weight in texturing for some images?

Thanks.

Hi AntoninPelly,
this seems like an issue with wrong alignment. Were your inputs georeferenced? Have many control points have you used? Were you able to solve those double walls or you processed the textures over them anyway?

Hi Ondrej, thanks for your reply,

Each scan position is georeferenced and set as exact.
I have used 3/4 CPs per side (square building) to solve the biggest gaps, but it seems like some of it remains in some places. It results in displacement where images “sticks” to scans and places where it seems to have no effect.
One problem is that my subject is a very high tower which make it difficult to place CPs on ground scan images with a lower quality.
I have indeed processed textures on very thin double walls that I didn’t notice.

Another Question not so related,

Is there a preferable order to import and merge scan and drone? Is it better to first import and align images and then import scans and align again or is it better to import everything together for the first alignment?

Are the images also georeferenced? If so, it is RTK, drone GPS or measured GCPs?
As the CPs are probably placed on the bottom part of the tower, there could be some misalignment on upper parts. On how many LSPs (laser scans) are those CPs placed?
As it is a high tower, you also need some points in upper site to get good alignment.
You didn’t answer the question about textures from laser scans. Have you turned laser scans off for texturing?
Regarding to your second question, basically it doesn’t matter. It could be done in one step, if you want to save some memory/time, you can merge components. But the alignment is based on the features between data, georeferencing and control points.

Images are originally georeferenced with the drone GPS, but the prior pose for drone images has been set to unknown at the beginning of the project.

For every ten or so drone images I am only able to place the CPs in 3 or 4 TLS images

Texturing for scans haven’t been turned off.

Misalignment on the upper part of the building Is not a problem as we also have scan positions on the roof and different floors.

According to your workflow it seems right. We just recommend to turn off the texturing from laser scans.
But to get a good model the slight misalignment should be solved, too.
Are there any errors on CPs in 1Ds view? If so, how big?