I warm this up again
Yesterday I did a long planned experimental approach on bottom vision.
I changed the rmod-fork so that I have a new data field in the tapes, telling me that I need alignment help with the bottom cam.
I can heck that "on" to any tape I like to have bottom vision alignment. I will use that only on bigger parts with more than 4 pins because it takes its time.
After pickup I drive to the cam, lower the nozzle and start a rectangle detection.
I have a red background on the image, a piece of foam with red paper on the nozzle holder, that makes it easier.
With a few filters, I get a really stable image of the QFP44 package I tested with. The vision system gives me the measured middle point as well as the angle, that is used for correction.
Its quite simple, I got it placed perfectly on the pcb
Hooray..
I see a few problems with smaller parts, as there is the nozzle visible in the background, that will need some work.
It will also need a few optimization runs as it tries to detect the rectangle, maybe a few tries with different threshold values to adapt to different parts.
I can also use different filter sets for specific parts, that is already selectable in the tapes, every part can have its own filters. To edit these I would add a function to pick a part and hold it over the bottom cam so I can adjsut the filter settings.
Then the machine needs to bring it back in its place.
Soon I have 16 days off, christmas
I will work on that and give you the results if it works.