loading
 
Defect Rubber
Giuseppe from Italy  [1 posts]
2 years
Hi Steven, I can not point out the defects in the central part of the rubber.
I also have problems with the form crop at the end of the program.
I have attached some pictures and program.
Best Regards
Giuseppe

  

program.robo
Steven Gentner from United States  [1370 posts] 2 years
Giuseppe,

First off it looks like the focus on those images is a bit off. You may want to tweak the Dlink camera lens to get the actual defects in view.

Second, which parts of the tire do you consider defects? Perhaps you can annotate the image to  better identify which parts are good versus not. Since a tire has a lot of texture on it is not clear if the horizontal side marks are defects or just part of normal tire operation.

Finally, do you have control over the final environment? Sometimes lighting from the side of the tire can help create better contrast between defects and actual. I.e. your current lighting is entirely dependent on the color of the defects. Lighting from the side will be affected by the peaks and valleys of the defects which may show up better on the camera.

STeven.
Giuseppe from Italy  [4 posts] 2 years
Hi Steven, I'm sorry for the previous images.You send images taken by a camera firewire. Defects are craters, large and small on the tire.
Best Regards
Giuseppe

   
Giuseppe from Italy  [4 posts] 2 years
Hi Steven, definitely the illumination will be made in the sides of the tire.
Will the cameras firewire with high resolution.
I attach some pictures with defects circled.
Best Regards

     
Giuseppe from Italy  [4 posts] 2 years
Hi Steven, I wonder if you could look at the pictures that I sent you and if there is a method to detect defects on the tires. I need to urgently develop the vision application that will be connected to a robotic system.
Best Regards
Giuseppe
Steven Gentner from United States  [1370 posts] 2 years
Giuseppe,

Unfortunately we don't have a solution for your images. There is some additional experimentation needed to get the lighting correct. A good rule of thumb is that if a human cannot tell where the defects are in the images its very unlikely an automated system can.

You would need to continue to play around with the lighting direction and see if you can create an image that is easier to see the defects on. We attempted a couple techniques on your images but cannot come up with anything worthwhile. Its a problem that is going to take quite a few days (if not weeks) to solve unless the image can be improved.

There might also be a way using a visible laser line to effectively scan in a 3D depth map of a tire as it spins. This would generate a lot more information that would contain the defects. The problem then would be to correctly process that depth information looking for irregularities (i.e. cuts and depressions against the known pattern).

A quick check on google shows a that the laser technique might be the way to go.

In other words a lot more experimentation and research is needed.

I would also recommend creating a more reliable relationship with a vision company if you have projects with deadlines as this forum is monitored only periodically and when we have time to respond. We make no guarantees of success nor are responsible to provide any solutions to the problems posted here.

STeven.

This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index