Width measurement

[Alexandre]

Hi friends

 

I'm starting a project where a steel strip width will be under measurement. This strip is under movement at 200m/s and its width can change between 700mm till 1900mm.

I'm acquiring a Basler camera, model Line Scan 4096 pixels and a 28mm lens to get measurement 2000mm far away the strip.

I have doubts about lighting and how the Line Scan image will be displayed in the VisionBuilder Assistant.

If somebody knows a literature to indicate, please inform me.

 

p.s. look my application attached

 

Thanks,

 

Alexandre.

[hooovahh]

You are right to be worried about lighting.  It is more of an art form and having something as open as this type of setup you will need to experiment alot to see what can be done.

 

In the past I have seen a line scan camera where a light bar was directed right at the lens of the camera.  This caused a saturation of all other light in the environment.  I'm not saying to use this product but the light bar looked similar to this.  This setup was for detecting objects falling and worked well because it was the only place where light was obstructed.  

 

In your setup a line scan camera coupled with a light bar might work as well, if you can mount some kind of light behind the steel strip.  Then you'll see light where there is no steel, and you'll see darkness where there is steel.  You'll then need to perform a binary threshold using some kind of tolerances or other image filtering, and then perform edge detection with a few degrees of rotation, and then some measurement for horizontal distance.

 

Sounds fun, and doable just some time experimenting with setups to get it right.

[Tim_S]

Anything you can do to reduce ambient light and just use the light source you've set up will help. Ask the camera and light supplier to bring in different models and types of lights; my experience is they carry around a bag of different light sources if not a couple of different cameras as well.

[EricLarsen]

Check out colored lights and lens filters as well.  Narrow band filters matched to light sources are a great way to eliminate ambient variations.  Consistent, repeatable lighting is the key to any vision application.

 

Edit:  Google "machine vision lighting" for dozens of companies who specialize in just the lighting you're looking for.

Edited March 27, 2014 by EricLarsen

[ShaunR]

I think you're going to have problems with a camera in this setup.

You have a bar for mounting the camera, so I assume the camera will be facing the sheet and drum?

 

You are trying to measure the width of a reflective material (steel) with a reflective (although grubby) background.. What tends to happen here is that increasing light intensity also causes flaring on the background. The background is moving, so the the reflection is not constant and liable to change with position and dirt. So it's not just a case of saturating it with light, you have to be really pedantic with the light levels and the discolouration on the drum itself will have you pulling your hair out  (it's a line scanner, so you have no context). Once you do get it set up, you will walk in the next day and it won't work properly anymore because someone cleaned/changed the drum and the lighting require re-adjustment. Your measurement system will also deteriorate over time as the discolouration changes on the drum as you will have to effectively "tune" it for the particular environmental condition on the day..

 

You are much better off mounting it in the gap between rollers then the light will work well for you (no discoloured, moving or reflective,  background) and will not be at all sensitive to variation as  the focus of the lens acts as a light filter. Stuff further way will be blurred and darker and require much larger changes in your source illumination to have any effect on your image background at all and you can krank the contrast up to find the edges. You can also choose a lens that has a very short focal depth and you will be basically checking against a pitch-black background. You wouldn't even need to have a homogeneous light source, it would work and be robust with ambient light.

 

That's the "ideal"........

 

If you cannot do that, the next best thing is to point the camera from the bar to the gap. The issue here, though, is that with your current intentions, you will be a lot further away, will have overhead lights back-lighting your image and perspective distortion. To mitigate the back-lighting you can mount it similarly as you have currently, but near the top roller pointing down. Since you are using a line-scan camera, you can fairly easily compensate for the perspective distortion with very simple (and not CPU intensive) calibration, but depending on what accuracy you want to achieve, you could also get away without it.

Edited March 27, 2014 by ShaunR

[Alexandre]

Thanks ShaunR for your comment.

I'm considering your tips. But the place where I'm installing the system changed. Look the attached. What do you thing?

I have no chance to measure the strip out the drum because the strip shakes if no drum under side.

 

Alexandre.

[Jordan Kuehn]

Will they let you modify the drum?  I've had some customers employ black spray paint in some places to help things out and were able to add it to their maintenance schedule.

[Tim_S]

What do the objects look like to your camera? All cameras pick up more infrared than our eyes do. Some cameras also pick up ultraviolet. Have you tried looking at the greyscale image? The camera may show the distinction between the roller and belt quite clearly where our eyes do not.

[Alexandre]

Yes Jordan, I can modify the color of the drum. But I'm trying a good contrast with the actual color. I hope get it.

 

Hi Tim, I've got the greyscale image and I'm trying a good contrast adjusting the controls in the NI Vision. But I'm using a webcam just to get a test until my official camera arrive (Linescan Basler 6144 pixels mono / Lens)