Hi,
I just started with my first iRVision project and after few days of trying I don't know how to go on.
In our factory we produce TV's and what we'd like to do is to automatically place and tighten screws on TV. Because of production process and variety of models produced on each production line, it's almost impossible to have fixed position of TV. That's why we decided to try camera vision.
I really precisely set up tool frame (also checked with CAD) and after all tests(rotations and so on) it seems OK.
But I'm not able to get good results from screwdriver offsetting. Each production line can produce tv's from 32" to 60", bigger tv is better for me, because it can't be moved or rotated on production line, but small ones can. So I thought that I should take a picture from above to find raw position of TV, than I could go closer to each screwing position, find a hole and place a screw. I also think if it's possible to have only two vision processes - "FindTV" (for each TV model...) and "FindHole" which will be used all the time. Is this good idea or would you suggest something else? In a fact any hint, experience or guide will be very usefull and appreciate.
Thanks, Tomas
iRVision setup troubles
-
Kabi -
March 18, 2013 at 8:28 PM -
Thread is marked as Resolved.
-
-
Hello again,
after another set of tests I found that maybe 2D Multiview setup could be helpful. Does anybody have experience with it together with robot mounted camera? I was able to teach both camera views, but when I try to set ref. pos. it processes both camera views at one time, so one of them is always wrong because robot is still on first position.
Running on R30iB controller
Tomas -
Kabi
I would try to do a "ruff" find to identify and locate the TV, and you should be able to utilize the Multi if you want.
Then have search positions offset by Ruff find VR to do a FINE find of each hole.
Then you can use the vision offset from FINE find directly applied to screwing point (search position, run FINE find, direct to screw, repeat)PS. both of these vision processes I would utilize Depalletizing to give me scaling. If you TV is well fixtured the Ruff could be depall, and the Fine not be depall.
doc
-
it processes both camera views at one time ?????
Running the processes it's up to you. Run one, save it on VR1,then run another and save it on VR2
-
Problem solved. When I tried to teach whole vision process from laptop it was impossible to snap first picture, move robot, snap second picture and then teach ref. pos. All I had to do was run each process from program as Fabian said and then use Set_Reference command. Anywas does anybody have experience with new possibility of standard digital camera usage with R30iB controller? Because higher resolution of this cameras could solve my problem.
-
The have a camera with 1296 x 966. Is it what you are using?
-
No, actually I use standard camera-VGA 640x480. But I will test camera you talk about next week. I hope results will be better. And one information to everyone: we received VCC camera instead of Sony, because Fanuc had problems with Sony cameras after tsunami in Japan. That's information from local Fanuc office. But these cameras works correctly only with R30iA controller, with R30iB this VCC camera has problem with exposition time, it is unstable. I spoke with serviceman from Fanuc and he confirmed this trouble and told me that only solution is to use Sony XC56 camera. So be careful in this case.