Polygonizing Cloud makes it shorter by 3-4mm

[[Ja...]]

Hi Team,

I had been experiencing some issues where the scanned items are coming up as 3-4mm shorter than my physical measurements. I had made some test to test this theory. This doesn't happen all the time but it has been happening for a long time occasionally. We rely on measurements using GOM Suite but we sometimes don't get the best accurate results. 

Here are my results:

This spool is coming up ad 5998mm physical check. We then scan this and process them through Cyclone as we are using an RTC360 scanner by Leica. 

The photo above is the "Point Cloud" in another software, it seems to shrink by 1 - 2mm every time we import it somewhere.

However, each time we polygonise the point cloud with a setting " 1, 3, 5" it seems to shrink even more for 3-4mm. Is there anything I could do to avoid this from happening?

Am I doing something wrong during the process? Is there a setting that I could check to lessen "shrinkage"?  Please let me know

 

Cheers,

James

 

[[Ja...]]

P.S I also have more photos with different types of spools - I am sadly still getting a similar results

[[Ch...]]

How exactly are you computing the length measurement in Inspect? This software has been tested inside and out to verify the integrity of any measurements and in my experience, any issues like this has typically been user error.

It looks like you've done a point-point distance, but used two planes? This could cause issues for a few reasons a few being:

1. The planes are not guaranteed to be parallel 
2. If the plane was calculated using a gaussian best fit, it can be more sensitive to outliers

To properly compare measurements using a point to point distance, you should first set up a local coordinate system using features on the part and then make sure you use the same points to take the measurements, then report them using the appropriate axis of the coordinate system.

I would create a Chebyshev outer on one side of the part, and a few points on the other, then use a projected point distance to simulate a height gauge check. 

 

Even better would be to create two planes as you have done and use one as a datum to check the true position of the other.

 

If you are still getting errors this large then let us know what else you have done.

[[Ja...]]

Hi mate,

thank you so much for your response. We tried different types of method to on how to measure these spools. We had a manual point to point (From end to end). I also tried a plane at the end and a cylinder in the middle to create an intersection points on both sides. I also make sure that the plane is not overlapping any MESH, ensuring that I have the flattest plane i could get. I also tried to use outer caliper which all of these, they all have the same results on the same spool and very similar results on other spools. It's always coming up short when it's a cotton spool.

Outer Caliper:

Plane to Plane (Point to Point): 

Different Software measurement: 

 

I also just spoke to a Zeiss representative and given me the same methods I had been trying - I'm just at a lost here as it has been an issue for years. I was advised to use GOM's auto Meshing, my MESH was from solid MESH to static MESH using this method. I was also told to use it with "Normals" which made no difference when creating a MESH. 
I've played with the MESH settings and no matter what I do, it's still coming up short.

My thoughts are perhaps GOM is considering the 2-3mm edge of my point-cloud as "Noise" is there a way for me to tell GOM to MESH manually - ? is there a setting for it that I could disable? 

Thanks

[[Ja...]]

Just a guess but this might be a data aquisition issue rather than how inspect is polygonising or measuring.  This is a very long object so im curious how you are aquiring the data and steps to keep volumetric accuracy under control.

In the same direction i could believe that a measuring tape will have a high level of uncertainty over these kind of distances .

[[Ja...]]

Hi James,

Thank you for your reply. We use a Leica RTC 360 scanner, and most of our scans are quite accurate. However, we’ve noticed that our straight cotton spools occasionally appear shorter than expected, despite having accurate measurements for spools with more complex shapes, like elbows.

To improve the quality of our scans, we spray the flanges and reflective metal with a chalk spray, which helps reduce noise and minimize extraneous dots. I also have professional pipe fitters with me to take physical measurements of the spool lengths.

I’ve tried various measuring tools to ensure the accuracy of these physical measurements, and they all yield consistent results. I’m at a loss as to why the scanned lengths are coming up short, and I’m more focused on understanding the underlying issue than simply finding a solution.

Thank you so much for throwing in what your thoughts are.

[[Be...]]

One hint from my side: If you have concerns regarding the polygonization step or the prostprocessing steps then you can simply import all of your native data as pointcloud into ZEISS Inspect. The caliper functionality could also be applied to point clouds. If the result of your caliper still shows a value less than your expected value you have a "problem" with your native measurement data or your physical specimen is really shorter than expected.

Hope this helps investigating your measurement results?

Regards

[[Jo...]]

Hello all, 

I'm not sure if it is the same problem that James, but with me it's happening something similar. 

I'm using a uCMM from BRUKER, the parts that I measure are small (around 30 mm) and have smaller details ( ~1 mm). 

The simple task of replace the mesh (G3d file) it recalculate and give me 2-3 microns diference at some results.

With the exactly same mesh file, that give me some different results. 

[[Be...]]

Usually this shouldn't occur, but different results could also depend on the alignment strategy. There are some alignments, e.g. like the Prealignment which could supply slightly different results in different runs (it uses internally some heuristics to find a good starting point). Also the way how elements are constructed could have an influence.

But for a complete investigation you have to contact your local ZEISS support and you have to supply your particular project.

Regards