Sharp edges in Mesh Appear Jagged and Rough

[[Mi...]]

Hi there, 

 

This question pertains to GOM 2022 inspect and the ATOS 5 blue-light scanner. 

I'm scanning a part with a very sharp feature line.  The goal is to use this scan in the future to cut to on a CNC machine.  The sharp edge looks okay until I zoom in--then it becomes a jagged mess of irregular triangles.  I've tried every combo of post-processing settings I can think of, and I still get some degree of the same issue.

Is this something that can be remedied by scanning technique, or postprocessing of the mesh?  And is there any way to "sharpen" the edges of the mesh, so to speak?  Or is this just the nature of scanning a very sharp edge, and this is as good as it gets?

 

Thanks in advance.

[[Je...]]

Hey Michael,

What scale are we talking about here? So which MV is used on the ATOS 5 and what order of magnitude does this sharp edge have and how precisely do you want it to be resolved? Furthermore, the quality settings or scanning parameters are also important to know.

With this knowledge, a better representation of the edge may be possible.

Gruß, Jens

[[Mi...]]

Hey Jens,

MV I'm told is 700, using a medium lens.  I used a couple thin layers of Spot Check to eliminate glare.  The camera is set on High Quality and Fast Scan, and the Post-Processing is set on Standard. This was the best result out of 24 possible setting configurations that we could find. 

As far as order of magnitude and scale, this is subtle feature line on a punch die, used to stamp a hood panel for an automobile, so this is not "sharp" as in a knife blade or something similar, it's a "sharp line" that you would see on car.  I don't know how to precisely measure that, I would guess the "tip" of the line is less than 0.25 MM.

How I would like it to be resolved ideally would be through post-processing, so we wouldn't have to re-scan.  Someone here thought there was a way to highlight the problem areas and perform some kind of action within ATOS to make the triangles smaller or something like this.     

[[Ma...]]

Hey Michael,

why did you use Fast scan instead of Full resolution? Is it a manual or automated measuring system?  The relative sensor position to the edge also plays a role. 
You can still avoid the triple scan points. However, the MV 700 is only partially suitable for resolving sharp edges. A smaller MV for these areas would be preferable.

What you mean by post-processing is actually the Mesh Editing workspace. There you can refine the mesh and edit it differently. For example, you could cut out the area and then fill the area to the left and right of the bridge with a bridge that runs exactly along the edge.
But be careful, this is no longer real data.

Greetings Manuel

 

[[Mi...]]

 

Hey Manuel,

We tried every possible combination of Camera Settings (High-Quality/More Points and Fast-Scan/Full-Resolution) and Polygonization Post-Processing Settings (Standard, More Details, Less Details, Smallest Data Volume, Low- Resolution, Adaptive). We ended up with 24 different meshes. The example I posted was one of the better results, and that happened to be the one with "High Quality and Fast-Scan" settings on the camera, and the post-processing set to "Standard". 

 

We're using a manual system. 

 

Thanks for the tips!  We'll have to try measuring with a smaller MV if possible.

Edited April 29

[[Ja...]]

Just to add here beyond what others have said, even if you use a smaller mv this edge will never be perfectly sharp like a CAD model, there will always be some tesselation effects .  If your downstream process really requires a CAD like edge then your route is in direction of reverse engineering.

Hope you find a solution that fits

R

X

[[Ma...]]

Michael currently I am being tasked to inspect blisks that have very sharp (<=0.003" edge radii) blades.  We use ATOS 5 airfoil with MV400, 270, 170, and 100.  I found through trial and error to capture real data on these edges that the MV170 set has sufficient point density to really capture the geometry.  Larger FOV such as 270 and 400 were not as sufficient for smaller details. 

[[No...]]

What you experience is a mixture of the natural flaws of the scanning process on one side and the way polygonization works on the other.

Imagine a theoretical part with an infinitely sharp edge, something that can't exist in reality. The sensor can only detect points which reflect light from the projector back to the camera. But what would happen on an infinitely sharp edge? If you think hard about it, there would be no points that sit EXACTLY on the edge. All of them would either belong to one or the other of the surfaces that form the edge. If you imagine scanning a perfect edge, then also none of the points gathered would end up EXACTLY on the edge. That means there would always be a remaining gap between the adjacent surfaces, however small it might be. Thus you would always end up with some amount of radius after polygonization, since the points nearest to the edge have no neighbours on the same surface they could connect (triangulate) to and so have to be connected to points on the other surface, resulting in a rounded edge.

The reason why the polygonized round edge is also jagged, lies in the fact that the scanned points near the edge are of course spread more or less chaotic instead of perfectly lined up parallel to the (real) edge. This is caused by many factors, but the most  obvious is noise.