POP3 - dependency of accuracy on scanner orientation

I have been generally happy with the volumetric accuracy of POP3 which has been mostly within 0.1 mm which is the figure stated in the spec. That’s until recently when I tried to use the scanner oriented vertically. The accuracy is substantially poorer.

The object is a cylinder about 34 mm in diameter and 100 mm in length. It is placed vertically on the stock turn table and some features are placed around the bottom to assist tracking. The distance is in the good / excellent range. Calibration is done just before scanning. Several scans done.

Set up with the scanner oriented horizontally :

Set up with the scanner oriented vertically :

Error in mm measured in Fusion 360. +ve values means the dimension of the model is larger than actual

May I know if this is something expected ? I have also tried to scan a cube and got the same result - the accuracy is poorer with the scanner mounted vertically.

Hi @new23d
A completed model need to be scanned with 3 angles , there is possible shift because the primitive don’t have specific features allowing the algorithms to get the proper alignment .

Some objects surface scanning better in virtual position, others in horizontal .
That why you should scan both directions and check out the results .

Is the distance the same at the base as on the end of the scanned volume ?
Distance between the projector and the surface ? Any changes in distance will give you different results as well since you measure it in microns , it is very crucial , even 50-100 mm shift in distance will already provide bigger errors , especially in microns .

Think about the pattern projected on that cylinder, they will diffuse on the edges .

I propose you make test on more volumetric object with specific features in place of primitives .

1 Like

Actually I have tried the same test on this not-so-primitive cube which has got some features on all faces to assist tracking and frame stitching but the result is the same, ie, the accuracy is always poorer when the scanner is mounted vertically.

However, your reply does bring out a point which has never came across my mind - pattern projected on that cylinder diffuses on the edges. As can be seen in this picture, the frame captured with the scanner mounted vertically has a lot less shadow. At first I thought that was good because more data means more accuracy , right ?

Now I realized that the additional data captured with the scanner mounted vertically are over areas where the light pattern is projected at relative shallow angles hence not accurate. Can the addition of these low-quality data contribute to the overall degradation in the accuracy of the model ? not sure but it sounds logical to me.


That is accurate observation @new23d
Some objects I have to move vertically, some horizontally, especially if they have round edges or under cuts .
From design the scanner is made to be scanned in horizontal position to the object at 45 degrees angle , so if you set it vertically you still need to think about the 45 degrees ( horizontal) other way you getting parallel projected lines what capture less depth and undercuts .

Horizontal or vertical position of the scanner really don’t matters , what matters are the object and it’s position .
The most accuracy and precision you get when the projected lines are sharpest , the fall off areas will be always bad , as that is where the lines get diffused at the end , scanning multiple angles eliminates the fall off areas , that why it is important to scan the object at least 3 angles and best in one session.

Don’t forget to have some fun !