PushPullFeeder Vision

[Jake]

Hello,

After setting up my liteplacer with OpenPNP and making sure it works properly in a default state, I started to make some additions. One of those was to add a line of semi-automatic feeders.

Digging around a bit, I decided on the Ploopy feeders, as they seemed easy to assemble and mass 3D print, while also being functional. 

I set some up and they work as described when pushed by my finger or jogged by the nozzle manually, so I tried to set up a ReferencePushPullFeeder to really test them out.

One thing about the ploopy feeders is that the tape covers the sprocket holes, making them quite hard to see. To combat this, I added a bit of red paper behind so that I could use "ColorKeyed" Vision type and hopefully see them better. 

I messed around with the pipeline and got what I thought were some pretty good lone circles defined for the sprocket holes, but It still fails to recognize them. I added a picture of the pipeline results and what the camera sees normally, as well as my machine.xml.

I have no idea how machine vision works in truth, am I going about this the right way?

Thank you,

Jake

[Mike Menci]

Jake - have you tried with default vision and the SMT tape ON TOP of your Ploopy feeder (all tape holes fully visible- does it work) ?

[Jake]

Hey Mike,

I did a few tests. One with tape on top of ploopy with plastic covering, one without the plastic covering, and one with plastic covering but a red piece of paper behind the holes. 

I attached screenshots of the results. The only one to be detected was the one with the red paper.

Jake

[Jake]

Here's another test of an unpeeled tape in the ploopy feeder.

The pipeline results show that all the circles are detected, yet auto setup fails to recognize sprocket holes.

A bit confusing if I do say so, but maybe my complete lack of experience is showing.

Jake

[Jan]

Hi Jake!
The internals use the expected distance between sproket holes to reject
false detections. That only works if the feeders Z is correctly
configured. From your last picture I'd say that the detection works well
which makes me assume, that the Z configuration is off.

Jan

On 24.07.2023 19:27, Jake wrote:
> Here's another test of an unpeeled tape in the ploopy feeder.
>
> The pipeline results show that all the circles are detected, yet auto
> setup fails to recognize sprocket holes.
>
> A bit confusing if I do say so, but maybe my complete lack of

> experiencePipeline results.PNG is showing.

> --
> You received this message because you are subscribed to the Google
> Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openpnp+u...@googlegroups.com
> <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Mike Menci]

Jan

Do you mean camera is not calibrated properly to see -what it should see?

[Jake]

Hey Jan,

I'm sorry, I don't know what you mean when you refer to Z configuration.

Jake

[Jan]

Vision operations require a Z level to convert pixels to mm/mil units. I
remember that sometime back Mark added some improvements to the feeder
to discard false detections. The only way he can do that, is by making
use of the known distance between sprocket holes.
IIRC vision operations and auto-setup requires a correctly configured Z
coordinate of the pick location.
Jake, go to the pick location and check if the Z value is good. (You
need to provide it later anyhow because auto setup only works on X and Y)

Jan

> https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com> <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer>>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openpnp+u...@googlegroups.com
> <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Jake]

Hey Jan,

I just checked and the Z level is right when the nozzle tip touches the tape. That is correct, right?

Jake

[Jan]

Hi Jake!
I just checked it on my machine and it seems, that the auto-setup
feature is more picky now: it only seems to work if there are more then
just the two sprocket holes detected that are close to the pick
location. From the last picture you shared, I'd assume, that if you
choose the lower part as pick location is shall work. If you choose the
center part, it probably does not because there are not enough holes
towards the top of the picture. If that does not work either, I hope
Mark has more ideas...

Jan

> https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com> <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com>> <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer> <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer <https://groups.google.com/d/msgid/openpnp/10702c04-3bb9-4c19-b89d-d7e3cefd9bc3n%40googlegroups.com?utm_medium=email&utm_source=footer>>>.

> >
> > --
> > You received this message because you are subscribed to the Google
> > Groups "OpenPnP" group.
> > To unsubscribe from this group and stop receiving emails from it,
> send
> > an email to openpnp+u...@googlegroups.com
> > <mailto:openpnp+u...@googlegroups.com>.
> > To view this discussion on the web visit
> >

> https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com> <https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com?utm_medium=email&utm_source=footer <https://groups.google.com/d/msgid/openpnp/a66a280b-11bb-4ee9-9261-33334595e1afn%40googlegroups.com?utm_medium=email&utm_source=footer>>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openpnp+u...@googlegroups.com
> <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/openpnp/53886006-11fe-465e-90e5-729ed8d0b8a7n%40googlegroups.com <https://groups.google.com/d/msgid/openpnp/53886006-11fe-465e-90e5-729ed8d0b8a7n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[mark maker]

Hi,

I'm on vacation so just a short email:

First, just to rule out the obvious, why are your sprocket holes on the right? Are the feeders at the back of the machine?

Circular symmetry vision is likely impeded because the sprocket holes are very close to that black edge and shadow. Straight edges are now explicitly rejected by the circular symmetry detection in order to reject the rectangular but similar sized part pockets. Maybe cou can change the 3D printed design to allow for a cutout that leaves ~0.5mm or so around the sprocket holes.

Key-color vision could still work with the edge present, but it needs a very vivid/saturated color. Yours is way too faded.

_Mark

[Jake]

Hey Mark and Jan,

Enjoy your vacation! Don't worry about my issue, It'll be figured out eventually.

They are at the back of my machine, mainly to keep all the tapes out of the way of people walking by, but if that's an issue I can move them.

Thank you for the insight into how the vision works! I'll try to implement a much more vivid color.

As for Jan, I'll also give that a shot now. 

Thank you,

Jake

[Jake]

Hey all,

I made a small 3D printed piece that fits under the tape in the Ploopy feeder, then got the brightest green spray paint from Home Depot I could find.

One thing about the ploopy feeder is that is doesn't completely remove the tape, instead it pushes it over the holes to make for easier cleanup, which I really like.

However, that has caused issues with machine vision.

I attached photos comparing absolutely no tape with tape covering the holes, and only the former worked. The pipelines dont seem drastically different though, so I'm wondering if theres any way I can make it work with the tape over the holes.

I would rather not have to make adjustments to completely remove the tape, but I will if I have to.

Besides that, the one without tape performed well in a practice job, except when it didnt see a part for some reason and push/pulled 2 times until it did.

Thank you,

Jake

[Mike Menci]

Did you ask the author if he has already a solution for vision ? 

@ https://hackaday.com/2020/01/23/a-ploopy-pick-and-place/

[Mike Menci]

I think you should be looking for vision adjustment to get something like you had already:

[mark maker]

That edge and shadow is (again) the problem.

Can't modify the model to cutout a section on the yellow dashed line?

Alternatively, for the color-key vision, you could perhaps improve the light to shine more from the left.

As the last resort, you could probably tweak the HSV Mask pipeline stage brightness and/or saturation.

If that does not work out, please send me a native image (with the cover tape still on):

https://github.com/openpnp/openpnp/wiki/FAQ#how-can-i-get-a-native-camera-image

_Mark

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/f46e2157-cd47-4537-9ce9-89bd2afa7e4cn%40googlegroups.com.

[Jake]

Hey Mark and Mike,

That's a good idea to reach out to the author, I'll try to shoot him an email.

I can try to remove the diffuser on the left side of the camera to get more light in the mix and see how that goes.

Also, I believe in the image I sent of the pipeline for tape over the holes I had tweaked the HSV Mask to get those results, with those being the best I could get.

However, I think I may just have the tape be fully stripped away to get better results and just have to deal with more cleanup. One of the main reasons making that cutout difficult is that it helps hold the tape down, so if i were to cut that segment,

the tape would be blocking the holes even worse. Even still, the tape makes the image quite blurry and I think its too much to try to get around.

I believe without the tape it should run pretty smooth, sorry for the hassle guys I just wanted to see if I could get it done with the tape. 

Thank you guys,

Jake

[mark maker]

OK with that latest shot, I no longer think it is possible. there are string irregular reflections in the cover tape.

But why don't you look at my all-3D-printed Push Pull feeder? This one is proven to work with OpenPnP and supports 0402 too.

https://makr.zone/new-all-3d-printed-tapereel-feeder/399/

 _Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/079d7161-d1e3-4d04-8c22-b5417dcf93b1n%40googlegroups.com.

[Jake]

Hey Mark,

Sorry for the late response, haven't been at my desk in a bit.

I saw your feeders, but as my machine is set up currently and with its position in the office I don't believe I would be able to incorporate them smoothly. Also, I have already (foolishly) made quite a few arrangements for the ploopy feeders.

I think I have a solution that's not too annoying, but thank you, you are always a massive help.

Thank you,

Jake

[Mike Menci]

General Question regarding transparent tapes: Which color of Filament would be the best to use under transparent tapes ? what color filament should i get?

[mark maker]

Something dark and as matte as possible for Circular Symmetry vision. I'm confident that the same deep vivid key-color, that is recommended for non-transparent tapes, is also fine for transparent tapes.

https://github.com/openpnp/openpnp/wiki/DetectCircularSymmetry#referencepushpullfeeder

But if you really want to go for the optimum, I guess that would be a matte black. Remember: all you need to reprint is the tape inset, so you can use a different filament just for that.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/5a39f1af-135f-4bfa-8918-a8cbae1658e1n%40googlegroups.com.

[M. Mencinger]

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/eb79c70d-10ce-851d-430e-28b8e3edb90b%40makr.zone.

--

Mike

"--

 **************************************************************************************************************

This e-mail may be confidential and it may also be legally privileged. If you are not the addressee you may not copy, forward, disclose or use any part of it. If you have received this message in error, please delete it and all copies from your system and notify the sender immediately by return e-mail. Internet communications cannot be guaranteed to be timely, secure, error or virus-free. The sender does not accept liability for any errors or omissions. 

 SAVE THE NATURE - SAVE PAPER - THINK BEFORE YOU PRINT!  "