Discussion about transformation deviation

[[Le...]]

Hi,

In measurement/scan, why the transformation deviation shown on the bottom of the picture is not equal to the average of all reference points transformation deviation. 

But the results are equal in reference point measurement, does anyone knows what is the difference and what is real defination of the transformation deviation?

Thanks, 

Leaf

[[Le...]]

[[Ja...]]

The answer is simple and complex at the same time.  The transformation of each measurement uses reference points AND scan data .  In reference point only mdasurements - no scan data - > numbers tie up

In the end the metrics are better assessed at a full project level if you are trying to understand how well things are transformed. 

The individual values in a measurement are not so helpful most of the time.

If you explain your reason for the question there may be easier wats to look at if.

[[Le...]]

Please sign in to view this quote.

Thank you for the help.

I was wondering whether the transformation works as you say when I choose measurement transformation by reference points only. Transformation by best-fit is based on scan data, is transformation by reference points only related with scan data?

 

[[Ja...]]

Hi,  this could get complicated quite quickly - in the end the software will use reference points and surfaces for adjusting the fine alignment regardless of what you choose here in this box.   This is something like the the initial allowable transformation method.

 I think it easier to talk about the root of your question.  Why is this interesting?

Is there a transformation issue in a project or similar?

[[Le...]]

A reference measurement data is required as ground truth in a measurement project that we have done researches on recently.  The distance between the centers of two referecne points is now we chosen as the reference data. But here came a problem that which measurement gets the more accurate distance, the measurement/Scan or measurement/reference point measurement. We noticed the difference mentioned above, and so asked the quetions.

[[Ja...]]

I think you need to be careful.

If you are looking to understand a bias to a 'known'then the reference cannot be determined by the system under test .

One thing you could consider is to include something like a guage block in your measurement project - this may give a base from which to work . 

I would suggest a good starting point is looking at repeatability of your mesh and your photogrammetry (if you have it )

I think a discussion with your distributor would be a good idea as this topic of accuracy /repeatability is indeed a deep and complex one.  

[[Le...]]

Thank you very much