Marble slab

[Hatef.fouladi]

indeed what you done is converting color image into grayscale by using color plane extraction or casting this is not real monochrome you could not solve color filter bayer problem in such image you have to use monochrome light and also it is better to use monochrome camera

[desalando]

Hello Hatef,

The Customer wants color images (to show his customers his marble slabs). Therefore we decided for color camera. It was not my first choice. I know that with monochrome camera we would have better results.

But his is the situation at the moment:
- LED white lighting (6500K)
- Color linescan camera

- Running conveyor belt as you can rebember

- At the moment dark marble slab with white areas. But there will be white marbles too.

Did you try image processing (for example segmentation with gray morphology) on my images?

[Hatef.fouladi]

hello again desalando
by my experience none off segmentation and grey morphology will be work well in this project because of multi color spectra in both back ground and object
here we need to texture analyse beside such function that you indicate
I am busy now but I promise work on your project in weekend just send me some video like you send me before and your line scan camera data
I try to find you a way that are more reliable but still claim that you could make your project easy cake just of you could change your vision device method

[Hatef.fouladi]

send the files to my email to prevent creating spams here

[BruceAmmons]

I would recommend balancing the colors for your camera.  The green seems much stronger than the other colors.  You can scan a sheet of white paper and adjust it until the color is white.

The rock paper looks interesting, but it looks like they just found the dark edges between the rocks.  That wouldn't help you here, though, because of the black and white sections of the marble.

I still think the only real solution is a 3D imaging camera.  If you keep the existing system for the color image, you could probably get a relatively low resolution, low cost version of a 3D camera, maybe a SICK Ranger D, and use that to get the profile of the marble.  With minimal processing you could line up the profile with the color image.  It really depends how accurate that profile edge has to be.

Bruce

[desalando]

Hello Hatef,

I send you some videos yesterday.

[desalando]

Hello Bruce,

I had also noticed that green seems much stronger than the other colors. I'll do white balancing as you suggested.

Since we can not change marble slaps, we will try to change background. To the that we are planning to change the position of the scanning line.

Camera scanning line at the moment:

We will move cabin, so at its new position camera scanning line will be between conveyor belt and roller conveyor.
So the background will be empty. With illumination on marble slab, we will see only marble slab. Other places will be completely dark.

[BruceAmmons]

That is a great solution.  It should work great with white marble.  I am a little concerned with the black marble, since the background will also be black.  Hopefully the black will be shiny enough to look different from the background.  Eliminating the background should make the processing much easier.  Post some pictures once you have moved the camera.

Bruce

[SimonAldred]

Yep getting rid of the conveyor belt is a good step. You could try to be clever here and use a red led light line below the marble for sizing and say a green led light line above and to the side for looking at texture in reflection mode if this is of interest. You could extract the the red and green images to give you twice the information from one camera. It all depends whether the green camera channel is totally isolated from the red light and the red channel isolated from the green light. 

[Nelson-Bridwell]

Would it make sense to place a long backlight below the conveyor gap so that you can detect the presence, position, and size of the marble slab as it rolls by?

Good luck!