Giuseppe,
What you are asking for is a bit involved, we can't just provide a 'quick example' and have that work. We work with companies to develop solutions that you are referring to that are appropriate for the product being inspected. For example, in your case, the blob inspection is probably NOT the right module to use. Instead, since this is a stitched product, I assume you are wanting to ensure that the black stitch around the lips is detected? It is probably best to use a straight image comparison between the two images in order to highlight the differences. But you still have a problem with alignment between the gold template and the test. Then you need to decide what amounts of differences would constitute and error.
Or perhaps its just the mouth area that would be checked? I.e. you want to ensure the black dot is in the right location?
Further, you need about 100 images of any particular product with different variations available to test a comparison on. Making assumptions about the product errors from just a few images will most likely fail to encompass all the ways in which the product can fail. This image set then forms the basis of a test suite of images that any technique would be tested against and verified to perform correctly. Otherwise when you change the parameters a bit you can never be sure if you didn't incorrectly break the detection for some images
Again, its quite a process and not just something that can be done quickly.
To answer your other question, yes, it is possible to add more images programatically to the blob inspection module (or OR module or shape module or ...). Just ensure that the 'monitor folder' setting in the training interface (ie. when you press the Train button) is on. This will cause a retrain whenever a new image appears/ changes or is deleted in that folder ... most likely from your own software. So not even the API is needed in this case.
STeven.
|
|
This Thread
