5 replies to this topic

[zuoli]

1.How vision inspection systems are calibrated or verified？

2.What is the principle of vision inspection system？

Who can help me?

[G M]

Generally you have to train vision systems in phases with good material, and distinct sets of items that you want it to reject.  It can vary widely from a few dozen samples up to thousands, depending on the complexity of what you're trying to train it to do.

 

Fundamentally it isn't very different than how you might train a person, but the samples have to be more formally grouped or identified for the machine to understand what you're presenting it with.

[GMO]

I've found in combination with the validation of "training" the system, there is also a test to check for ongoing efficacy.  What that test is will depend on what you're checking for.

 

I've seen some systems where a contaminant is on a wire, or a (food safe) but differently coloured item goes through.  Or a calibration "card".  Best idea is to talk to the suppliers.  They will normally advise.  

[zuoli]

Test sample？

• One confirmed good product
• One confirmed bad product, representing each defect type that the system is inspecting for
• One marginally defective product representing each defect type that the system is inspecting for

Generally you have to train vision systems in phases with good material, and distinct sets of items that you want it to reject.  It can vary widely from a few dozen samples up to thousands, depending on the complexity of what you're trying to train it to do.

 

Fundamentally it isn't very different than how you might train a person, but the samples have to be more formally grouped or identified for the machine to understand what you're presenting it with.

[Tony-C]

Hi zuoli,

 

Generally, vision systems are set up and checked in accordance with the manufacturer’s instructions or recommendations.

 

As per GMO’s post, you need to set up the equipment to detect and reject typical faults that you want removed. When I used vision equipment with glass bottling (inspection of bottles immediately prior to filling) we had about 15 test samples of bottles that we ‘programmed’ the equipment to reject. These were checked at start, hourly and end of the production run. Typical faults that were rejected were:

Glass fragments

Various foreign objects

Chipped containers

Dirty bottles

 

The ‘test’ samples were set up based on typical faults and data from complaint analysis to identify the most common customer complaints.

 

Kind regards,

 

Tony

[GMO]

It might help if you can share what kind of vision system or defect you're looking for?  But the equipment manufacturer themselves should be the most expert in this and be helping you.