accurate measurement with vision

[lightgoal]

Hello,

I'm just asked to make an accurate measrement for a part with height of 30mm. I must design a system that understands if the part is 30.01 or 29.99 (.01mm). and (I think) that means I must chose a camera that gives me one pixel for every .01 mm so I need a camera that hath at least 3000 pixels in one diamond. and this doesn`t sound like a good idea.

now I'm a little mixed up and I'm even not sure even I have enough knowledge to do this. Now I want some one to guide me do what or maybe read what.

thank you.

[BruceAmmons]

You would need a camera with a minimum resolution of 3000 pixels.  More would be even better.  This accuracy could be difficult to obtain without spending a fortune.

Another option you could consider is using two cameras.  Position one at each end of the part.  Once you determine the fixed distance between the cameras, you can measure very small offsets with each camera and get high resolution measurements at a much lower cost.

You could also use a line camera.  This would give you much higher resolution in the axis of measurement without costing a fortune.  Once again, it depends on the object you are measuring.

Bruce

[wbadry]

Bruce is always a friend in need

Is that true that to achieve 0.01 mm accuracy, you may use twice the expected accuracy (that means 6000 pixels for having precise measurements) . I'm sure it is not feasible , however, do u agree that it is the right choice if we walk by the book.

I may be against using 3000 pixels since this problem may evolve:

• Depath of field: if the object is 3000 pixels wide, it means mounting a camera requires that "all width" is utilised, this would may bring about "Distorted Image" due to camera lense- I suffered from that problem. 

Bruce speaks as expert. He knows it would cost a fourtune to get such a camera. Perhaps the line scan camera is the optimum solution for this problem like he proposed. Two cameras is something I personally may suffer in terms of setup and accurate offsets.

[BruceAmmons]

I agree that distortion would also be a major issue when using a single camera.  You would have to locate the camera several feet away from the part to minimize distortion, and even then you might have issues with a thick part.

I still like the idea of two cameras mounted 30 mm apart.  As long as you don't need any information about the center of the part, you can get very accurate measurements at each end with practically no distortion.  It is not that difficult to mount two cameras with parallel axis, as long as you do a little planning.  Calibrating the spacing is not too difficult either, as long as you have a part with a known length.  You could even use cheap telecentric lenses (the Computar TEC-55 comes to mind) for zero distortion.

Bruce

[wbadry]

That's another valuable info. Thanks Bruce and hope this all can be implemented to serve our friend who is in the real problem.

[lightgoal]

thank yoiu for taking part to solve this problem,

I think what bruce says is a great solution, but I want to ask him what if I fix one side of the part in a definite place and measure the other side, I mean is it a good idea to put the camera at the above 5mm of the part and say that we add the measured distance to 25mm and we know that the other unmeasured part is absolutely 25. the only problem is that we must find a way to make sure that the part is placed correctly under camera. maybe a pneumatic jack can come usefull here.

and about a 3000 pixel cam, I said at least 3000 pixel and you're absolutely right. I just wanted to show that it's somehow so difficult.

any other idea and is warmly appreciated.

thanks

[BruceAmmons]

You can measure both sides if you want to.  All you need to do is figure out the distance from A (any point, really) on camera 1 and B on camera 2.  Make all your measurements relative to A on camera 1, and B on camera 2.  The total measurement is the sum of the three measurements.

You could do it with one camera like you say if you were absolutely sure of the position of the other end of the part.  I don't think you can be that certain down to 0.01 mm, though.  If nothing else, the surface you would be pushing it against could slowly wear.  It doesn't take long to wear off 0.01 mm of a surface.  I would feel much more comfortable measuring both ends.

By the way, having some idea of what kind of part you are measuring would be helpful.  A picture of it would be even more useful.  Some of these ideas could be useless when related to your part.  Other ideas could also be suggested after seeing the part.

Bruce