On Thu, May 12, 2022 at 1:16 AM Rafael Skodlar <ra...@linwin.com> wrote:

> Hi Andrew
>
> Tensorflow came to mind right away. It's in the machine learning (ML)
> class. Amazing software if you ask me.
>

You don't need machine learning to do this.  All you need to do is an RGB
histogram.  If the oject is red then the sum of the red pixel values will
be higher than the sum of the green pixels.    You can do what they call
"classic" computer vision in openCV.  If the parts are always in a fixed
location just grabe the pixes at that location or even use a lens that
looks at where the part will be.


> https://www.tensorflow.org/install
>
> Obviously you'll need to do some research to connect things together.
> There is a tensorflow application for Android which I used on my PDA
> (phone is one app on it) to detect objects. It was fun to direct the
> camera towards a TV and detect moving objects on it.
>
> I would start with RaspberryPi, add camera (even a USB kind), and
> install tensorflow. That way you could start experimenting with ML to
> detect different chips, count them, and kick off the bad ones from the
> conveyor belt. A RaspberryPi with a transistor activating an
> electromagnet could do the job. Estimated cost? $60-80 I think.
>
> Be careful with handling low voltage RaspberryPi and more powerful side
> to handle the mechanics.
>
> See examples of what can be done
> https://www.tensorflow.org/lite/examples
>
> Came to mind, it would be good to have a machine to sort different kinds
> of screws from a "collection box". US screws mixed with metric ones is a
> total nightmare for me.
>
> Best wishes,
>
> Rafael Skodlar
>
>
> _______________________________________________
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>


-- 

Chris Albertson
Redondo Beach, California

_______________________________________________
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to