Re: Broken Marker!

[Rob Spanton]

Hey Andy,

On Wed, 2016-03-23 at 19:48 +0000, Andrew Barrett-Sprot wrote:
> Due to some quick experimenting at the Southampton doing today, I've found
> out that *Arena marker #23 is broken, *as in *it literally can't be seen by
> robots. *I've tried regenerating it from kokimarker. As I'm stuck for other
> things to try I thought I'd open it up to the  mailing list.

I have just tried viewing marker 23 with libkoki.  It does detect it.  It seems
to have a higher failure rate, but it does detect it successfully.

I don't have time to look into this thoroughly at the moment, but it looks like
it is because a chunk of the large black square gets turned into a white square
in the adaptive thresholding stage (see attached).

The correct mailing list would be srobo-devel for this.

Cheers,

Rob

[Andrew Barrett-Sprot]

Hi Rob,

It's an issue that is directly affecting teams, I've tried viewing the marker with both types of cameras and in different lighting conditions and neither could see the marker at all.

I think this is urgent enough to warrant a proper look, if you're too busy could you point me in the direction of getting the debug information out of the robots themselves?

Thanks,

 Andy B-S

--
You received this message because you are subscribed to the Google Groups "Student Robotics Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to srobo-devel...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Rob Spanton]

On Wed, 2016-03-23 at 21:38 +0000, Andrew Barrett-Sprot wrote:
> It's an issue that is directly affecting teams, I've tried viewing the
> marker with both types of cameras and in different lighting conditions
> and neither could see the marker at all.

I'm not seeing any teams having an issue with it.  This marker will have
had this feature for the last 4 years, and throughout those years it
hasn't caused any fatal issues.

It would be nice to improve the performance of the vision system so that
it didn't have this problem, but right now there are quite a few other
priorities.  This isn't a problem that will be solved in a couple of
minutes either, it is one that requires careful analysis etc.

> I think this is urgent enough to warrant a proper look, if you're too
> busy could you point me in the direction of getting the debug
> information out of the robots themselves?

You don't need to pull any debug information from a robot.  libkoki will
run on any normal Linux machine.

Cheers,

Rob

[Peter Law]

Hi Andy,

Andy wrote:
> It's an issue that is directly affecting teams, I've tried viewing the
> marker with both types of cameras and in different lighting conditions and
> neither could see the marker at all.

If this issue is affecting teams there is a really trivial thing they
can do to avoid it: use a different marker for testing. There are 27
other arena markers they could pick for general testing, and three
other arena sides they could use!

My own testing (running on my desktop and admittedly at fairly short
range) shows no particular difficultly with this particular marker.
Things I suggest you check for include:
- focus of the camera
- damage to the marker printout

> I think this is urgent enough to warrant a proper look, if you're too busy
> could you point me in the direction of getting the debug information out of
> the robots themselves?

As Rob notes libkoki runs on any linux box (for reference I'm using my
desktop with my own webcam -- not even an SR issue camera); if you
want a hand setting it up and playing with it I'm happy to guide you
through that.

Thanks,
Peter