Hello everybody,
I post my question in this new thread, even if I found some resemblance with my problem in a previous one called “Pattern Matching Comparison” and in “Pattern Matching Strategies” (tutorial document” .
I'm working on a vision application with "poor an non uniform lighting" and "occlusion or overlapping parts". I'm using Labview 8.2 and with a GigE Dalsa M1020 camera. During my tests the camera supplier showed me how the Dalsa "Sapera LT" works. It seems that it works well, with a "edge based search". I could verify "as much well" because a couldn't compare it in parallel with NI Software, but actually this is not my problem, because I'm an NI developer and I want to find a solution with the NI platform.
With the same application I have to recognize very different objects (embossed plastic plate/label different in shape and color) situated on a belt conveyor.
When I started, I have based my program on "Pattern Matching" algorithm. It worked quite well with some objects, but with others the identification percentage was very (too) low, so I have modified my program introducing, also, "Geometric Pattern Matching" (different search based on the object type). Using this way the identification percentage is quite good, but I have that the program finds the exact pattern position but with a 180° rotated angle. This is a big problem for me because I have to pick the objects with a robotic arm and place its in a mould for successive working.
I guess to understand that this problem is related to the template image that doesn't contain enough features but I don't know what I have to do to improve my template image.
I attach some images and Vision Assistant script.
Does someone give me a suggestion to solve my problem?
Thank you in advance.
Best regards.
Michele
AUROS
Software e Tecnologie per Automazione
