working distance

[lightgoal]

how can I make an accurate measurement from an object that does not have a constant working distance to lens of camera. the object may be placed in a surface that is 10 cm far from cam but in different places on surface which camera takes image in defferent angle each time. and another problem is that what if even we don't have a predefined working distance to the object and the distance between lens and object is changed in each image.

I need to make accurate measurements in length and width of the object in accuracy of 0.01mm.

thank you

[pallen]

0.01mm is pretty small.  To even have a hope of measuring that distance, it needs to be at least two pixels in your image. 

If you have that kind of resolution, then non-linear calibration can help with measuring at different distances.  Although if your angle is changing to, that could present other problems.

Good luck. 

[Kristen_K.]

Hi lightgoal,

Pallen makes two really great points about resolution and using the Non-Linear Calibration example.  Thanks Pallen!

Kristen H.

[lightgoal]

Message Edited by lightgoal on 04-28-2009 01:08 PM

[lightgoal]

hello,

sorry for empty messages that was mistakely put there!

I appreciate for taking part solving this problem. and i got the idea of using calibration software. but the problem is not totally solved yet. the problem is that the total surface which holds the object under inspectioin (buttons in mentioned example) moves in front of the cam and so the cam takes picture in different distance in each time. and that means the buttons will have one size in each picture. and i want to make accurate measurements of the object.

now i'm looking for an idea maybe a technic or any thing that can solve this problem.

thank you

[Robert_Eastlund]

We usually solve this type of problem by using a telecentric lens, see http://www.edmundoptics.com/TechSupport/DisplayArticle.cfm?articleid=239.  As long as the part remains within the depth of field of the lens, you will be able to make accurate measurements.  The difficulty with telecentric lenses is that if the object is very large, then the telecentric lens will also need to be very large, and the cost of the telecentric lens is primarily driven by its size.  How large is your field of view?

[muks]

---

We usually solve this type of problem by using a telecentric lens

---

I beg to differ. In one sentence, 0.01mm is 99.9% not adivisible to try. If you can share your application then we might suggest some indirect method.

[1100fps]

The Automate show is in Chicago next week, this will have (I assume) the largest concentration of vendors and suppliers to the machine vision industry.  You could walk the show floor and discuss the issue with a wide range of suppliers.  NI used to have a booth there every year.  Sadly, no more, so there is not much to help with the software side, but there are lots of optics, and lighting solutions that you could investigate.  Telecentric lens may be part of a solution; remote focus lenses are now becoming common. These can be lighting fast. Looking to find the distance between two objects to an accuracy of 10 microns is a big ask.  Just agreeing on where the edge of an object is to within 10 microns can be tricky.  Most edge tools rely on subpixel interpolation to conclude the location of an edge, but that requires consistent angle and illumination.  Another reason to bring your problem to the Automate show.