Defect detection in pasta production

[Tom Theron]

I am consulting for a pasta factory. They have a problem with cracked pasta being packaged, instead of rejected. If a flat view of the pasta on a conveyor belt is taken, the sticks are normally cylindrical "pipes" , but when cracked , normally split into halves, they appear as shadowy trenches. What is the possibility of using machine vision to pick up this variation? One must keep in mind that the pieces are randomly scattered on the conveyeor, thus I would imgagine the only possible method would be to do a spectrum analysis on the color variation, and to raise an alarm when the shadow spectrum crosses a set threshold.

Has anybody implemented something similar?

[ryanm]

Hi Tom,

If the cracked pasta is noticeably darker, Machine Vision should have no problem distinguishing the good and bad pasta. For maximum speeds, you may want to consider using a B&W camera, because processing time on B&W images is much faster. In addition, if you can do some filtering and identify bad pasta by intensity instead of a pattern match with rotation, this will yield faster performance.

Ryan M.
National Instruments

[Kyle_V]

Hi Tom. Do you have any images you can upload to the thread? We might be able to suggest some vision algorithms and methods.

Thanks!

Kyle V

[Tom Theron]

Hi Ryan,

I requested samples for testing, which I'll receive next week. I will post the images when I have them.

Thanks
Tom

[Tom Theron]

Kyle,

I couldn't get hold of photos of the cracked macaroni process yet. This is of another process, where spaghetti is processed. The photos is a series with different percentages of defect, as indicated by the photo filenames.

I'll post photos of the macaroni process soon.

Thanks
Tom

[Rick Saunders]

Hi,

I am the Vision Application Specialist here at NI Corporate. I would be glad to discuss your app over the phone. Feel free to give me a call on my direct line. (512)683-0033.