I am inspecting dimension of large object (above 300mm)

[31238]

There are two choices:
One is using a short focal length lens to get a wide enough FOV. But this way may induce perspective distortion.
The other is using a long focal length lens to get small FOVs but magnified details. Then integrate small FOVs to get a dimension across different FOVs. This way may induce a lens distortion due to close shot distance. Also this way may affect dimension accuracy due to overlap and combination many times.
The question is which method I should take and what the strong and weak points are.
Tks in advance.

[BruceAmmons]

How far away can you place your camera? If you can get a good distance between the camera and the part, the distortion can be corrected using NI's calibration routines.

Another possibility is a telecentric lens, which has practically no distortion. I don't think they make them that big, though, so you would still have to move the camera or the object using a stage. The telecentric lens does give you a precisely calibrated image, and if you have a precision stage you can move exactly one image width at a time. I built a system that measures within a few microns using composite images assembled from multiple images.

Depending on the type of measurement you are doing, it might be possible to set up two cameras at opposite ends of the part and ignore t
he center of the part. You would need to figure out the distance between the cameras, but once you did that you could measure the distance between points in the two images.

Bruce

[31238]

You said "if you get a good distance..., the distortion can be corrected using NI's cablibration routines."
What's the range of that good distance? In addition, In NI's support documents, NI never tells how the dimension accuracy will be gotten after spatial calibration. Can you kindly make me aware?

[BruceAmmons]

It is an approximate term. Good distance depends on your application and the accuracy required. If you can position your camera so it is two or more FOV lengths away from the object, your distortion will be minimal. The acceptable amount of distortion is determined by your application.

The dimensional accuracy of the NI calibration routines depend on the grid used for calibration and the amount of distortion. Once again, this can't be predicted very well.

Bruce