How to define the aquisition resolution?

[cmunikat]

Hello all!

I am aquiring images for a machine vision system. When I wondered about which resolution should I use, I thought that any value that the vision algorithm works is a good resolution value.

However, I imagine that there is some related theory, perhaps related to nyquist sampling theorem?

Please comment.

Thank you for your response!!!

 

[cmunikat]

Hello!!!

I've found at least one constraint for the minimum resolution.

The resolution is the same as the sampling frequency, i.e. the frequency at which we sample the real object, for example in pixels per unit distance. As it is known, the minimum sampling frequency in order not to have alliasing is twice the maximum space frequency contained in the real object.

If the resolution is not enough, alliasing and thus distortion occurrs. One may think that this is not a problem since it is a systemic distortion. However, the objects that the machine vision system has to see may vary slightly in spatial frequency, what will cause that two similar objects will be displayed as very different ones.

This effect is bound to cause problems in most machine vision systems, what do you think? At least, I think that it affects mine's.

Please comment.
Cmunikat

 

[watertreader]

I think you might have to be concerned about magnification and also the pixel size of your camera system(assuming you are working on digital camera system).

By nyquist sampling, the smallest object feature that is imaged onto the image plane(or the camera CCD) must be sampled by at least 2 pixel of the CCD or better. The smallest feature must be demagnified, considering the magnification of the vision system

Also depending on what your camera vision system require. The resolution requirement might not be too critical.

Maybe you can look up site like https://www.edmundoptics.com or some camera manufacturer website. I believe they would have more information.

 

[cmunikat]

Watertreader,

Thank you for your response.

What I imagine from what you say is that if I see the camera as a black box which gives has images as a result, and if I see that each object detail appears at least in 2 pixels of the result image, then I am sampling with a correct frequency according to the sampling theorem.

From this point of view, the magnification and pixel size are transparent, and I can check if the frequency is correct by looking the result image.

Do you agree? please comment if this is what you mention in your response, and if I am loosing something...


Thanks!

 

[watertreader]

There are multiple parameters in an imaging system which would determine whether you can see an image(assuming you mean by eye, looking at an image).

I think I have mentioned about magnification and resolution etc... there is a whole lot more of parameters you would have to consider.... depending on your needs...they could be important or not important. Also good contrast is also required for differentiation.

Assuming the other parameter does not really affect your acquisition, Nyquist theorem state that the smallest resolvable image must be acquired by at least 2 pixels(one should go for more...but might increase storage burden). The size of the smallest image is in turn determined also by the magnification

Hope it helps!

Added after 5 hours 1 minutes:

Perhaps you can also check out this book "Digital image processing" by kenneth R castleman. There is one chapter discussing on optics and system with regards to resolution but from a microscope system viewpoint. So there might be quite a number of sections in this chapter that might not be totally applicable.

I guess in a machine vision viewpoint...perhaps the resolution criteria might not be so stringent as compared to microscope(hope I am right here)...are you trying to design the optics or the camera system?

Hope it helps!

 

[cmunikat]

I just need to make a proper selection of the camera and optics.