Unfold function – distortion/gaps in the unfolded cloud

Feel free to ask any question here
Post Reply
celamk
Posts: 12
Joined: Sun Sep 24, 2023 5:08 pm

Unfold function – distortion/gaps in the unfolded cloud

Post by celamk »

Hi!

While unfolding a point cloud along a polyline, I ran into a problem with the dimensions of the resulting raster.

My test cloud is about 2000 m long and is a scan of a high-voltage power line running through fairly flat terrain. There are a few ditches and small depressions, but these are the largest elevation differences along the entire test cloud. To create the unfolding line, I first used the Trace Polyline tool, but later I switched to creating a vector in QGIS and then importing it into CloudCompare.

I created two polylines:
  • a 2D only vector (no Z coordinates), which later I had to align manually with the cloud I wanted to unfold,
  • a 3D vector, draped over a digital terrain model that reflects the terrain breaks and depressions (this one fit the cloud perfectly).
Then I ran the unfolding algorithm. The results are shown in the attached images (unfolding along the 3D vector in the top half, along the 2D line in the bottom half). Section 1 shows firs hundreds of meters (raster unfolds towards right side of the image), and there is no shift there. In section 2, the mentioned elevation changes are visible, and they produce gaps in the unfolded cloud. This is particularly clear in the top half of the image. In the next image (section 3), the length differences caused by accumulated error are already so significant that they raise questions about whether this data is suitable for measurement purposes.

This leaves me with a few questions:
  • When using 3D profiles, does CloudCompare “stretch” the scan (viewed as a cross-section) so that the distances correspond to the length of the vector?
  • Should I expect even larger differences when scanning mountainous terrain?
  • Is there a risk that a vector closely following the cloud could cause the cloud to be “normalized” and displayed as a flat surface, in addition to stretching the scan?
  • In mountainous terrain, where a 2D line would cross (pierce) the terrain surface multiple times, could this cause deformations in the unfolded cloud, or should it not affect it?
Thanks in advance for any suggestions!
Attachments
unfold_error_s3.jpg
unfold_error_s3.jpg (343.35 KiB) Viewed 12331 times
unfold_error_s2.jpg
unfold_error_s2.jpg (458.62 KiB) Viewed 12331 times
unfold_error_s1.jpg
unfold_error_s1.jpg (327.17 KiB) Viewed 12331 times
daniel
Site Admin
Posts: 8247
Joined: Wed Oct 13, 2010 7:34 am
Location: Grenoble, France
Contact:

Re: Unfold function – distortion/gaps in the unfolded cloud

Post by daniel »

Hi,

1) Yes, the X coordinate of the unfolded cloud will be the curvilinear abscissa (arc length). So it can virtually stretch the resulting cloud.

2) It all depends on your supporting polyline. To me it makes sense to keep it 2D, so that you don't induce any scaling.

3) If the projection direction is Z, and the polyline is 2D (XY), then you should not see any flattening (as the height, which is the distance from the points to the XY plane, should remain correct).

4) Same as 3, with these conditions it should not affect it.

If you don't want to align the cloud, you can load the 3D vector, and then multiply (Edit > Scale/multiply) the vertices by 0 along the Z dimension. This way it will be flattened, but keeping the right X and Y coordinates.
Daniel, CloudCompare admin
daniel
Site Admin
Posts: 8247
Joined: Wed Oct 13, 2010 7:34 am
Location: Grenoble, France
Contact:

Re: Unfold function – distortion/gaps in the unfolded cloud

Post by daniel »

Hi,

1) Yes, the X coordinate of the unfolded cloud will be the curvilinear abscissa (arc length). So it can virtually stretch the resulting cloud.

2) It all depends on your supporting polyline. To me it makes sense to keep it 2D, so that you don't induce any scaling.

3) If the projection direction is Z, and the polyline is 2D (XY), then you should not see any flattening (as the height, which is the distance from the points to the XY plane, should remain correct).

4) Same as 3, with these conditions it should not affect it.

If you don't want to align the cloud, you can load the 3D vector, and then multiply (Edit > Scale/multiply) the vertices by 0 along the Z dimension. This way it will be flattened, but keeping the right X and Y coordinates.
Daniel, CloudCompare admin
celamk
Posts: 12
Joined: Sun Sep 24, 2023 5:08 pm

Re: Unfold function – distortion/gaps in the unfolded cloud

Post by celamk »

Thanks for clarification Daniel!

Turns out that 2d polygons are not only safer option but faster to make as well :) Double win :D
Post Reply