Einschneideabweichung - What is it?

[[Iv...]]

+++Repost from old Forum+++

Hello everyone,

although I'm using the GOM ARAMIS 12M Adjustable for quite a while, I sometimes still wonder about the calibration procedure and it's outcome. 

For example: How can it be, that although I calibrated my system quite carefully (calibration plate mounted on a tripod) the 'Einschneideabweichung' is 

in the range of 0.07, instead the targeted range below 0.03. 

Or in other words: What does the 'Einschneideabweichung' actually represent? In which (theoretical) scenario would it be zero? Maybe a detailled explanation 

of the theoretical background would help me already to improve my calibration process.

Greetings,

[[Iv...]]

Hi,

the intersection deviation or in German "Einschneideabweichung" is the value of the validity for your calibration, as you mentioned.

Maybe the way the software calculates the value would help you to understand what exactly the intersection deviation means.
1. The software takes a facet from the left camera picture (primary camera).
2. The software is searching for the same facet in the right camera picture.
3. The software calculates the 3D-coordinate point (the middle point of the facet). Let's call this point the real 3D-Coordinate.
4. The software knows because of the calibration and so the epipolar line where this 3D-coordinate should be. Let's call this point the theoretical or ideal 3D-Coordinate.

5. The software compares both middle points. The position difference between the real one and the theoretical/ideal one is the intersection deviation. The unit of the difference you get is in pixel.

The reason why you can't get the 0 as a value is simple. Because in reality it is not possible to obtain any exactly same position twice. You always have a small deviation in rotation, height and more. 

Greetings,
Ivan

[[Ol...]]

in addition to Ivan's post:

as we us a model to describe the lens distortions we will always have some deviations between the real distortion of each lens and the model of the distortion we use in our calculation. The deviations are quite small, below 1/10 of a Pixel (pixel size is in your case 3.5µm).

You mention a target range for the intersection deviation. Especially for the Adjustable system this is only a general recommendation and can be higher for some setups. This not necessarily means that these systems have a lower measurement quality. The intersection error depends on the camera type, lens type and calibration object type.