Hi guys,

While evaluating an application with a customer, I had a question related to our wrist camera and the feature/part detection.  The  machine tending application was using our camera to locate the part at a different stage of the milling process and bring it to the next stage.  The customer was wondering if the new features just machined or chips left over the part by the CNC would affect the part detection.  Thanks to the threshold feature in the part recognition module, those type of situation will not stop the part detection.  Lowering the threshold value will allow the detection of the part even if some new feature is added and even if some are hidden.  Goal is to avoid 100% of the part surface/contour to be hidden.  But a threshold value to low, might end up having known contour mixed up if a similar contour is created in a different area...  

Have you been facing similar issues in your applications, so far?  How did you manage it?

@mikebeauchemin it's possible to adjust your threshold to get a better reliability but in the end if you get a bunch of chips that randomly form a shape that match your objet you will always end-up with problem.

I've seen a nice workaround for this from one integrator. They had to get a air line on the robot, they have one nozzle blasting air on the camera lens to keep it clean from cutting fluid and chips. They use the same thing to blast the workspace with air to clean the place before using the vision. Sometimes the vision will need some outside help like this ;)

@mikebeauchemin on important thing I forgot, you can add new features it won't change the recognition of your part. As long as the original shape of the part is still present.

In short, if you teach a square bracket and add some holes in it, it's fine. If your square bracket becomes a circle, the vision system won't find the square anymore, I'm sure you get the logic of that !