Today I continued to play with scanning thin objects which is kind of challenging. The object being scanned is a plastic plate 100 mm long and 2.75 mm thick. The set up is shown in the picture, marker tracking used. 3 scans done. The distance is in the excellent / good range. What’s interesting is that the error at the top of the plate , which is closer to the scanner, is much larger than that at the bottom. After a number of exercises, I thought I have got some basic understanding on how to obtain accurate scans but now I am not so sure. The rules seems to be more complicated than I have thought. Would like to hear some feedbacks.
You can’t have excellent distance on top and below , 3D structured light scanner have only one excellent distance when scanning type of object you have here so total proper behavior .
3D scanners was invented to scan difficult to recreate objects and shapes and primitives are not one of them .
But if you have fun learning how things works , that’s all what matters .
There is not poorer accuracy at closer distance . Wrong usage of the scanner are the cause of the results .
I know 3D scanner is not needed for modelling a rectangular plastic strip but I am interested in knowing how to get maximum accuracy out of the scanner under different scenarios. A plastic strip is used in this exercise just because it’s easy to measure and cost nothing. Ball bars are also very “primitive” but scanners are able to scan them with superb accuracy so I doubt the object being primitive is the reason.
Anyway, I have made a simple modifications to the object and retried - paint the base black. Now the result is reversed, the thickness at the top becomes very accurate. That at the bottom is poorer but it’s not too bad considering the fact that it’s further away from the depth cameras.
Still don’t understand why it is so.
Are you still using all recommended fusion and meshing settings? I would recommend trying some things manually, and definitely try standard fusion just for a comparison. And get the newest Revoscan. Something is decidedly different with Advanced fusion.
At least record the settings used and include them in these posts, since if it’s recommended settings all these comparisons are likely at different point distances and meshing levels depending on how much total volume you’re capturing (it will include those marker dots in the volume, for instance, unless I’m mistaken).
You can also try eliminating everything but the object in keyframe edit mode before fusing. And if using recommended settings, I would reload the project after saving the edits since I’m not sure what actions trigger a re-evaluation of the recommended settings.
One last thing, that base black paint looks like it might cause some reflections.
You don’t get me , to have the exact accuracy the object need to be at the same distance through it’s length relatively to the scanner .
If top of the object is close than the base it will have different accuracy even with 20 mm change of distance .
Hi TheBoatScans , I always use advanced fusion. The recommended parameter values are always used. The point distance is about 0.35 mm. It seems that the recommended value depends on the distance range. The closer the object is, the finer the projected light pattern on the subject , hence higher resolution and shorter point distance becomes meaningful. Isolation is done before meshing.
The newest version of Revoscan , ie 5.4.3 is used.
I am not aware of the key frame editing function but doing it manually for 200+ frames may be too tedious …
If you are just culling the garbage, you use Edit All Keyframes and you can remove all the dots (for example) at once. Beware it’s directly editing the depth images and there’s no going back.
1 Like
OK, just tried it. Removing the marker images in all the keyframes before fusing actually degrades the accuracy. For the scan chosen, the error in the thickness at the bottom of the plastic strip increased from 0.1 to 0.16 . The error at the top of the strip is almost unaffected and stays at 0.05. Interesting.
Tried again on a previously done scan with the base of the plastic strip in white. By deleting the base ( unfortunately also some markers ) from all keyframes followed by re-fusing and re-meshing, the magnitude of errors in the thickness of the strip at the top and bottom are swapped. Clearly the existence of the base has an effect on the error. May be it affects the stitching of keyframes during fusion in some ways ?
You have to remove everything under frame editing , and leaving only the object without anything else .
Don’t use recommended values for fusing, it will always decrease the accuracy to lower .
Use advanced mode and the best available fussing resolution .
When you edit frames , edit all at once , not each separately.
Get rid of the markets too . After editing click fusion tab and click save when window prompts.