Re: [Paparazzi-devel] RE: distance measurement for landing.

[Christophe De Wagter]

We use a mouse sensor together with a 10mm focus lens and a pitch gyro divided by GPS ground speed to get ground altitude. Cheap and increasingly accurate when closer to the ground.

(gps_groundspeed in m/s)  /  ((optic flow in radians per second) - (pitch rate in radians per second)) = +/- height / cos(roll)

see http://scholar.google.nl/scholar?q=Vision-Only+Control+of+a+Flapping+MAV+on+Mars+&hl=nl&btnG=Zoeken Equation 10 for a slightly more precise formulation.

Most surfaces have quite good contrast. Robbe even made a helicopter autopilot with this idea: http://www.google.nl/search?q=helicommand+3a

The difficulty lies is finding a lens that has the optimal properties: too narrow is too much flow (blur) and not enough light. Too wide is on the 18x18 pixel sensors of the mouse sensor not good either... on grass or on concrete different lenses are optimal. In other words: just like any vision based systems, you need good images. 

Good thing though: the mouse sensors have an image quality indicator: just neglect the measurement if it is bad and interpolate in beween the good measurements.

Christophe De Wagter
+31 6 28 02 94 08

On Fri, Nov 13, 2009 at 2:43 PM, gisela.noci <address@hidden> wrote:

Yep, that was precisely optical flow. If interested I have some papers on
the subject, describing a project where a fixed wing UAV was flown in a
canyon, and then between buildings on campus, not to a GPS track, but to a
'guideline' through the maize, with the optical-flow process keeping the UAV
from approaching objects. Problems were the optics - lenses - ,ccd
saturation, and used 4 'cameras' , one forward with some down pitch, one
down, one port, one starboard. Oh, and a 400mips DSP.. Heavy and required
reasonable bore-sighting between sensors. There is no simple, or rather, no
single, simple solution. But combining many simple solutions is the way to
go. There are some good UWB radar solutions out there - try McEwan
Technologies - look for Model PER-24 (WWW.getradar.com, I think) - a 24GHz
0.5m to 10m range radar, 50mmX50mmX12mm - it has great potential.

Joe

-----Original Message-----
From: paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
[mailto:paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
] On Behalf Of Gareth Roberts
Sent: Friday, November 13, 2009 1:20 PM
To: address@hidden
Subject: Re: [Paparazzi-devel] RE: distance measurement for landing.

Hmm, it's a British thing, we all learn about it as children.
The Dam Busters was a project to destroy 3 German dams during world war
2.  They did this using round bombs that looked like barrels, which were
spun so they skipped like stones to get over the anti-torpedo
nets...quite incredible, but they discovered they needed to be a certain
height above the water for it to work, so they had two spotlights (nose
and nail) angled inward so that at the right height the two beams would
converge, so the pilots knew exactly what height to fly at (18 metres in
a heavy bomber in the dark over water...).
http://en.wikipedia.org/wiki/Operation_Chastise

As for the Gumstix, I'm working on a project at the moment using a
BeagleBoard (similar idea, bit more powerful I think) and a couple of
webcams to do some stereo ranging/optical flow work on a helicopter.
I'm sure it's possible to lift all the stuff, if you are lifting a
powerful linux sbc we may as well run the whole stack on it, which I
think is an excellent idea anyway (especially with the availability of
RT arm linux kernels).  Make the tiny into a simple IO device instead.
It's just that its quite a complex layer of hardware to implement what
should be a fairly simple sensor.

There must be something out there people :).
I saw a fantastic paper at EMAV by a guy using optical mouse sensors for
altitude control and object avoidance.  From what I remember, the return
from the sensors depends both on how high you are and how fast you are
going.  Wonder how accurate that would be, because we already know how
fast we are going...

--G

On Fri, 2009-11-13 at 09:38 +0100, address@hidden wrote:
> On Thu, 12 Nov 2009 15:54:13 +0000, Gareth Roberts
> <address@hidden>
> wrote:
> > Catches with the laser?  You'd probably need to lift an SBC
> > (gumstix/beagleboard), although you might get away with a CMUCam.  Not
>
> Well, we already have Paparazzi running on a Gumstix Verdex and it
> does have a header to directly interface CCD-chips including working
> V4L-drivers.
>
> > sure of the range, especially in bright sunlight over non-reflective
> > surfaces, like grass.  The ham radio one looks more promising, but
> > difficult and heavy.  Incidentally, I looked at the laser range-finder
> > idea quite a while ago after I saw a documentary on the Dam Busters!
>
> Never heard of them. Do you have any pointers to this documentary?
>
> Marcus
>
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel




_______________________________________________
Paparazzi-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/paparazzi-devel