Testing OpenMV H7 as a 'flyby' camera

[@markgreco1962]

There is NO doubt in my mind you will make a pick and place machine. Seems to be a logical progression. Amazing work Chris as usual.

[@EnglishTurbines]

About 23 years ago, we invented a Laser Cutter with an optical recognition system on it. A basic Sony Camera, looking down onto the machine bed finding woven in squares in the fabric. It then used them to orientate and cut the part out...Laser cut Woven badges for Football Shirts was our first application. Company made a fortune selling machines and our Software....I just designed the Cutter....🤔😏😏😀🇬🇧

[@ParthBhat]

I remember subscribing to your channel 7 to 8 years ago and it’s amazing to see what all you’ve been doing. Thanks for this. Now coming to the video, I started working on my own pick and place machine as well, but the aim was to use esp32 (just obsessed over it), apparently. Neoden YY1 already have done this on stm32H7 with 2 cameras which is down and up facing, connected with the spi display and there’s a master board which is talking with cameras over uart. Super fancy thingy they have made without using a microprocessor to run the whole pick and place machine. Excited to see where you would be going ahead with this. But yes its definitions work for pick and place machine no doubt! Cheers! 🎉

[@李福生-j5q]

This is a perfect pick and place demo video for OpenMV H7 with global shutter. And we have already desiged all in one global shutter vision sensor with OpenMV cam H7 Plus also RS485 protocol and 2NPN outputs, it might be more convenient for your application .

[@-r-495]

controlling ambient light is essential. this is quite impressive for the size. I‘d want to dig in to optics and lighting first to reduce outside disturbances. Use a filter in the cameras optics and get the lighting right (soft/hard/spectrum) before heuristic noise reduction („clarity“). I‘ve spent weeks before industrial CV systems, thank you for sharing this interesting prototype, I see you have interesting requirements to meet but I understand the benefit of time myself. Hope to be back soon, I have found quite the rabbit hole. μPython is cool, compile it and it‘s reasonably fast.

[@iforce2d]

Yeah this light ring is just one I had sitting around. It's far too wide, and fixed color. For an actual PnP setup I would probably use RGB leds, and two smaller rings, and some diffusion. As for requirements I don't really need flyby, I'm just drawn to the technical challenge :)

[@paularthur7386]

Nice. We're playing around with the OpenMV H7 as a person detector. Its a lot of capability for not much money, and working well for us currently.

[@AlJay0032]

Very impressive how far you have come.

[@robindebreuil]

Amazing how fast cameras can operate these days - to think that people used to have to pose for over a minute :) Great experiments, looking forward to seeing the inevitable PnP!

[@Terry_Fritz]

It took me a while to work out what the function of the setup was. I guess the library of component shapes will be quite large for a commercial application.

[@iforce2d]

Yeah I should have showed an example of a machine that uses this method so we all know what it's for, was gonna edit in a little clip of that but I forgot.

[@BuzZ.]

back on track !

[@brucebaxter6923]

Thanks for doing something with the open mv. I found mine very powerful for the price.

[@iforce2d]

Yeah I'm tempted to get another, would prefer one the same as this but they're not selling them anymore. Would also be nice to buy just the board only since I already have a color sensor I don't need...

[@brucebaxter6923]

@@iforce2d ok. I didn’t know they are out of production. I have one that I’m not likely to use again.

[@avejst]

Impressive build Great job, as far as I can programme 🙂

[@jacobdavidcunningham1440]

hmm I don't understand this test rig, it's trying to have something fast moving? does circular make sense over linear motion? 7:11 ahh, I guess when it's zoomed in like that the motion doesn't matter just repeatable test stand

[@iforce2d]

Yeah in actual usage the movement would be linear, but this is a simple way to make a good-enough test setup.

[@GNARGNARHEAD]

great project, thanks for sharing 👍

[@PiefacePete46]

Has the quadrant reached full speed before the first hole passes the camera? I could not see anything that suggested otherwise, and it may not even be relevant. This is beyond this old brain's capability... but I love it! 👍

[@iforce2d]

It is relevant, the motion needs to already be at the expected constant speed for the first photo, at least if you want multiple photos and you want to make things easier by having the same sleep interval between photos. If you only have a single nozzle I guess it doesn't matter.

[@MrZeek101]

Look at you. Chris's AI electrical Arduino sweatshop great job Love where you're going with this. It save you a lot of work

[@sayedmuhammadidrees1453]

Great