Accurate 3D Measurement in 2D Images – Part II

Playing around with my 3D Stereo Vision profiles provides valuable insight into the difference between 2D machine vision and 3D machine vision.

Measurement in 3D

I have measured using stereo-vision the following values:

  • r1 and r2 – the stop cock radius
    • the values deviates with 0.02 mm
      • the pixel-resolution is 0.08 mm – indicates sub-pixel repeatability
  • d34 – the 3d distance between center of stopcock 3 and 4
    • the value is 15.16 mm
  • dy is the distance from stopcock 4 to top of the wide white ring
    • measuring the distance along the y-axis removes issues with locating features with different z

3D Measurements in 2D Images

Measurement in 2D

I have implemented the same measurement in 2D.

To make a comparison I have then done measurements setting z=5 mm and z=0 mm.

Measurement in 2D with z=5 mm

2D measurements with z=0 mm

We can do the following 2D measurement comparison:

  • Radius measurements have higher repeatability
    • There is a offset as a function of z – changes almost 0.07 mm equal to a pixel when z goes from 0 to 5 mm
    • Without measuring z – it will be difficult to know the distance
    • With stereo-vision the offset is gone and the system handles height deviations
  • Distance can only be measured introducing an offset to compensate the perspective
  • We also see that the longer cock to cock distance have a bigger offset due to the larger distance.
    • A 5 mm height difference accounts to 0.21 mm or a ratio of 0.04

The correct Z is 5 mm for the stop-cocks.

Conclusion:

The Scorpion 3D Stereo Vision is extremely valuable when measuring complex parts with natural height deviation.

It can also make the difference when you need to do measurements and you objects are too big to fit under a telecentric lens or you have to make a low-cost measurement system.

There is a great value in add 3D measurement techniques in 2D images with 3D Stereo Vision to assembly verification and measurement systems. You can make systems that you know is working correctly.

%d bloggers like this: