[Canberrauav] Todo list (software)
Stephen Dade stephen_dade at hotmail.com
Wed May 2 18:43:14 EST 2012
Hi everyone
We had a small meeting during the MHV meetup last night and made a list of
all the software modules we need to make:
. Georeferencing of detected Joes in an image
. Compression and transmission of Chameleon images to the ground station
. Checklist module (Chris and myself are working on this)
. "Fake Camera" for SITL testing
. Outback Challenge Rules Module - implementation of our D2 in code
. Interference testing of the 5.8 GHz and 900 MHz radio links
. Improved navigation code
. Analysis module for picking which Joe is the correct Joe from the 1000's and images
and 10's of "detected" Joes
. Joe image viewer for ground station
. Integration of ground station software (qGroundControl, APM mission planner, etc)
. GPS NMEA stream as per The Rules
. Caching offline maps for moving map display
. More bottle drops! And building up a predictive model of the ballistics.
If anyone wants to start work on any of these modules, send a reply back to
the list to we can sort out who's doing what.
Thanks
Stephen
[Canberrauav] Todo list (software)
Andrew Tridgell tridge at samba.org
Thu May 3 18:17:38 EST 2012
Hi Stephen,
Thanks for doing the list!
> . Georeferencing of detected Joes in an image
yep, not done yet (we do have some old code for this, but its not
integrated into MAVProxy for realtime use)
> . Compression and transmission of Chameleon images to the ground
> station
this is done now
> . Checklist module (Chris and myself are working on this)
great!
> . "Fake Camera" for SITL testing
not done
> . Outback Challenge Rules Module - implementation of our D2 in code
not done
> . Interference testing of the 5.8 GHz and 900 MHz radio links
Andrew Moss and I did this with some students from Dickson College, and
we found the 900MHz and 5.8GHz radios did not interfere. Unfortunately
we did the rest while using a 900MHz antenna on the 5.8GHz radio! (the
antennas look extremely similar unfortunately), so I'd like to redo it
some time.
> . Improved navigation code
yep, needed but not a top priority
> . Analysis module for picking which Joe is the correct Joe from the
> 1000's and images and 10's of "detected" Joes
yep, not done yet
> . Joe image viewer for ground station
this has been done now
> . Integration of ground station software (qGroundControl, APM
> mission planner, etc)
not done, although we can run them separately
> . GPS NMEA stream as per The Rules
not done yet
> . Caching offline maps for moving map display
not done yet
> . More bottle drops! And building up a predictive model of the
> ballistics.
yep!
I'm hoping to fly the Telemaster on Sunday with the realtime Joe
detection running. All welcome!
Cheers, Tridge
PS: Jack picked up the APM2s for the failsafe board from me today
Matthew Ridley dr.matt.ridley at gmail.com
Mon May 21 11:39:38 EST 2012
Hi all, I am back from my sojourns.
Is there an update to this task list ?
I guess we can refine it tomorrow night regardless.
Matt
On 3 May 2012 18:17, Andrew Tridgell wrote:
> Hi Stephen,
>
> Thanks for doing the list!
>
> > . Georeferencing of detected Joes in an image
>
> yep, not done yet (we do have some old code for this, but its not
> integrated into MAVProxy for realtime use)
>
> > . Compression and transmission of Chameleon images to the ground
> > station
>
> this is done now
>
> > . Checklist module (Chris and myself are working on this)
>
> great!
>
> > . "Fake Camera" for SITL testing
>
> not done
>
> > . Outback Challenge Rules Module - implementation of our D2 in
> code
>
> not done
>
> > . Interference testing of the 5.8 GHz and 900 MHz radio links
>
> Andrew Moss and I did this with some students from Dickson College, and
> we found the 900MHz and 5.8GHz radios did not interfere. Unfortunately
> we did the rest while using a 900MHz antenna on the 5.8GHz radio! (the
> antennas look extremely similar unfortunately), so I'd like to redo it
> some time.
>
> > . Improved navigation code
>
> yep, needed but not a top priority
>
> > . Analysis module for picking which Joe is the correct Joe from
> the
> > 1000's and images and 10's of "detected" Joes
>
> yep, not done yet
>
> > . Joe image viewer for ground station
>
> this has been done now
>
> > . Integration of ground station software (qGroundControl, APM
> > mission planner, etc)
>
> not done, although we can run them separately
>
> > . GPS NMEA stream as per The Rules
>
> not done yet
>
> > . Caching offline maps for moving map display
>
> not done yet
>
> > . More bottle drops! And building up a predictive model of the
> > ballistics.
>
> yep!
>
> I'm hoping to fly the Telemaster on Sunday with the realtime Joe
> detection running. All welcome!
>
> Cheers, Tridge
>
> PS: Jack picked up the APM2s for the failsafe board from me today
[Canberrauav] Todo list (software)
Andrew Tridgell tridge at samba.org
Mon May 21 13:37:28 EST 2012
Hi Matt,
> Hi all, I am back from my sojourns.
welcome back!
> Is there an update to this task list ?
there have been a lot of software changes. It's probably easiest for you
to look at the two main repos:
https://github.com/tridge/cuav
https://github.com/tridge/MAVProxy
plus spend some time together going through the changes and getting you
the data we captured on the weekend. Basic status is that we have a
system that can do the following all running in real-time on the panda:
- capture in 1280x960 8 bit with in-camera gamma and exposure control
- capturing reliably at 7.5 fps
- debayering at full-res
- automatic frame time determination using the image frame time header
- joe scanning while flying
- geo-referencing of joe while flying, including excluding hits based
on the search boundary
- jpeg compression on pandaboard for transmission to GCS
- display on GCS of a 'mosaic' of found Joe images, with click to display high
res image of individual joes along with geo-referenced location
The two pandaboard CPUs are basically flat out doing all of this. We're
now using the Neon SIMD vector extensions of the ARM CPU to do the jpeg
compression and some of the joe scanning.
The geo-referencing is terrible - its still using my original rough
geometry code. We need to replace that with proper rotation based code
and take into account the trapezoid shape of the image and the lens
distortion. Once I get you the data maybe you could help with that?
Basically we need to fix up pixel_position() in cuav/lib/cuav_util.py. I
was thinking of using the DCM code from
mavlink/pymavlink/examples/rotmat.py to do the rotations.
If you want to try it out on a Linux box, the following command will run
most of the critical code:
scantest.py camera/raw/*pgm --view --mosaic --mavlog flight.log --boundary CMAC-boundary.txt
(you need the data of course). If you left click in the mosaic window it
shows you that image. If you right click it switches to high-res
mode. Try scantest.py --help for more options.
> I guess we can refine it tomorrow night regardless.
did we plan a MHV meeting tomorrow? I do think we should have a software
meeting this week in preparation for this weekend. If not at MHV then
perhaps at my place?
Cheers, Tridge
Copyright 2012 http://canberrauav.org.au/