Feeder Vision Demo
123 views
Skip to first unread message
Jason von Nieda
Jul 18, 2012, 2:07:11 AM7/18/12
to ope...@googlegroups.com
Hey folks, I've just published a short video showing the new feeder
vision system. Check it out!
http://youtu.be/VviPLNRWDFk?hd=1
I suggest watching in full screen and HD so you can see the screen better.
Jason
vision system. Check it out!
http://youtu.be/VviPLNRWDFk?hd=1
I suggest watching in full screen and HD so you can see the screen better.
Jason
Bryan
Jul 18, 2012, 3:58:03 AM7/18/12
to ope...@googlegroups.com
Really, really cool Jason. Thanks!
Bryan.
Bryan.
...
g.lab...@glelectronics.nl
Jul 18, 2012, 10:58:11 AM7/18/12
to ope...@googlegroups.com
Good job, very impressing.
I wish I had some more time to help on openpnp.
Richard Spelling
Jul 19, 2012, 11:18:04 AM7/19/12
to ope...@googlegroups.com
really cool, Jason.
A suggestion if you are not already doing it.
Every tape seems to pull differently, depending on various factors.
Your vision system compensates for this.
Are you recording the error for that feeder and adjusting the pull
length for next time?
I.E, when you pick the part, you determine the positional error with the
camera, then record this, so the next time you pull that tape you know
you need to go .78mm longer to compensate for pin flex, etc.
--
Visit my online store for solar electronics: http://www.spellingbusiness.com
------------------------------------------------------------------------------
NOT sent from a Blackberry. Sent from a free hand-me-down PC running
free open source Ubuntu Linux... :-P
A suggestion if you are not already doing it.
Every tape seems to pull differently, depending on various factors.
Your vision system compensates for this.
Are you recording the error for that feeder and adjusting the pull
length for next time?
I.E, when you pick the part, you determine the positional error with the
camera, then record this, so the next time you pull that tape you know
you need to go .78mm longer to compensate for pin flex, etc.
Visit my online store for solar electronics: http://www.spellingbusiness.com
------------------------------------------------------------------------------
NOT sent from a Blackberry. Sent from a free hand-me-down PC running
free open source Ubuntu Linux... :-P
Richard Spelling
Jul 19, 2012, 11:19:45 AM7/19/12
to ope...@googlegroups.com
or if you are using feeders, you adjust the "probable part location" to
the last location, then look in the window from there.
On 07/18/2012 01:07 AM, Jason von Nieda wrote:
the last location, then look in the window from there.
On 07/18/2012 01:07 AM, Jason von Nieda wrote:
Jason von Nieda
Jul 19, 2012, 10:50:36 PM7/19/12
to ope...@googlegroups.com
Hi Richard,
The feeder stores a set of vision offsets. When it first starts, these
offsets are null, so the first time the feeder is commanded to feed
that part it performs the vision sequence twice. For reference, I call
this "front-loading the offsets".
So, we define the Vision Operation like this:
1. Move the camera to the defined pick location.
2. Do a template search within the area of interest.
3. Calculate a set of offsets from the pick location to the point
where the vision found the part.
Then, the first time the Feeder is commanded to feed since startup:
1. Perform the Vision Operation.
2. Subtract the calculated offsets from the feed start location and
move the pin to that location.
3. Subtract the calculated offsets from the feed end location and drag
the tape to that location.
4. Perform the Vision Operation.
On all subsequent feeds:
1. Subtract the calculated offsets from the feed start location and
move the pin to that location.
2. Subtract the calculated offsets from the feed end location and drag
the tape to that location.
3. Perform the Vision Operation.
The reason for front-loading the offsets on the first feed is that it
lets us, in normal cases, perform the vision at the end of the
operation. The benefit of this is that in most cases, when it's time
to feed a part, we will have just finished placing a part. If we had
to perform the vision first, and then feed, we would visit the part
location, feed, and then visit the part location again which would be
a wasted movement. So instead we always work off of the offsets that
were calculated during the last feed, right before the part is picked
up.
Hope that answers your question. If not, let me know.
Jason
> --
> You received this message because you are subscribed to the Google Groups "OpenPnP" group.
> To post to this group, send email to ope...@googlegroups.com.
> To unsubscribe from this group, send email to openpnp+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/openpnp?hl=en.
>
The feeder stores a set of vision offsets. When it first starts, these
offsets are null, so the first time the feeder is commanded to feed
that part it performs the vision sequence twice. For reference, I call
this "front-loading the offsets".
So, we define the Vision Operation like this:
1. Move the camera to the defined pick location.
2. Do a template search within the area of interest.
3. Calculate a set of offsets from the pick location to the point
where the vision found the part.
Then, the first time the Feeder is commanded to feed since startup:
1. Perform the Vision Operation.
2. Subtract the calculated offsets from the feed start location and
move the pin to that location.
3. Subtract the calculated offsets from the feed end location and drag
the tape to that location.
4. Perform the Vision Operation.
On all subsequent feeds:
1. Subtract the calculated offsets from the feed start location and
move the pin to that location.
2. Subtract the calculated offsets from the feed end location and drag
the tape to that location.
3. Perform the Vision Operation.
The reason for front-loading the offsets on the first feed is that it
lets us, in normal cases, perform the vision at the end of the
operation. The benefit of this is that in most cases, when it's time
to feed a part, we will have just finished placing a part. If we had
to perform the vision first, and then feed, we would visit the part
location, feed, and then visit the part location again which would be
a wasted movement. So instead we always work off of the offsets that
were calculated during the last feed, right before the part is picked
up.
Hope that answers your question. If not, let me know.
Jason
> You received this message because you are subscribed to the Google Groups "OpenPnP" group.
> To post to this group, send email to ope...@googlegroups.com.
> To unsubscribe from this group, send email to openpnp+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/openpnp?hl=en.
>
mcgyvr
Jul 25, 2012, 7:35:36 AM7/25/12
to ope...@googlegroups.com
Just to be "picky" ha ha ha you are feeding the tape backwards.. Holes should be on the right.
or if you are using feeders, you adjust the "probable part location" to
> the last location, then look in the window from there.
>
>
>
> On 07/18/2012 01:07 AM, Jason von Nieda wrote:
>> Hey folks, I've just published a short video showing the new feeder
>> vision system. Check it out!
>>
>> http://youtu.be/VviPLNRWDFk?hd=1
>>
>> I suggest watching in full screen and HD so you can see the screen better.
>>
>> Jason
>>
>
> --
> Visit my online store for solar electronics: http://www.spellingbusiness.com
> ------------------------------------------------------------------------------
> NOT sent from a Blackberry. Sent from a free hand-me-down PC running
> free open source Ubuntu Linux... :-P
>
>
> --
> You received this message because you are subscribed to the Google Groups "OpenPnP" group.
> To post to this group, send email to ope...@googlegroups.com.
> To unsubscribe from this group, send email to openpnp+unsubscribe@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages