On Sat, 5 Apr 2008 13:19:37 -0700 (PDT), Adam Chapman
<adam.chapman@student.manchester.ac.uk> wrote:
>I wanted a digital image to be processed in Matlab, which can aquire
>images over an IP protocol, and is excellent software for image
>processing because it is built around matrix algebra.
OK, so you're doing your computing in the airplane, not on the ground.
>Also I wanted a camera that output images in digital 'pixel' format
>because converiing a frame from line-tracing cameras into digital
>format is very slow.
There are many different types of cameras available. Most (not all)
have a "raw" format which outputs a bit mapped image. This is the
fastest mode because it involves no post processing (compression,
image enhancement, image stabilization, or color correction). You
won't find this mode in the cheap cameras, which output JPG's
directly. However, since you have a flying Matlab computah, you
probably have the horsepower to do your own rasterization in the
computah, which means you can use a cheap USB camera.
>In another project I am working on for the
>(http://www.challenge.mod.uk/) we tried using an analogue camera on an
>RC helicopter and the quality was awful- particularly we had problems
>with colour synchronisation and ghosting.
There's only one way that I know to get ghosting. That's if you use
NTSC/PAL video, and you have reflections on the interconnect cables
due to mismatched impedance terminations. Color smear is caused by
bad group delay on the same interconnect cables, where the different
frequency components of the video signal, have different (or changing)
phase shifts.
However, you're correct about analog cameras not being the right
choice. NSTC video only needs about 320x240 resolution to be useable.
No need for more dots for TV. However, if you're going to split
pixels with Matlab, you need all the dots on the screen as possible. I
suggest you look into using a hacked digital camera with a really big
CCD sensor.
>The aircraft in this project is for the ESO-UAS competition (http://
>http://www.hertfordshire-connect.com...ContentID=1234) where we are
>legally allowed a UAV weighing up to 20kg! We can't make it this heavy
>though because the wingspan would be ridiculous.
"Teams will be required to demonstrate four aspects of autonomous
flight, including image recognition, a pylon race against time,
aspects of navigation and being able to land the unmanned aircraft at
a designated spot."
Ouch. That's not going to be easy. I would be tempted to suggest big
and slow, so as to give the processors more time to make decisions.
>I wanted to put a mini-itx computer onboard to do the processing but
>the weight (~2kg) was not justified.
It's not just the weight of the board and cooling system, it's also
the weight of the power system. How many minutes/hours flight
duration is expected? I couldn't find anything in the terse press
release and the almost blank "Preliminary Design Specifications"?
>Basically I chose a camera that had a good digital quality and allowed
>tuning of parameters like frame rate, exposure time etc. I also liked
>the fact that it had a serial output, which could perhaps be used on a
>future aircraft to drive servos, allowing a pan/tilt system or even
>the aircraft flight control system!
Good logic but I wouldn't have done it quite the same way. If you're
going to have a mess of independent sensors, that all communicated to
a mess of dedicated computahs, each with their own particular
function, then the camera has to fit into the network communications
topology. If you're using TCP/IP for onboard communications, you
might want to look into an IP network camera.
>I'm only working towards an image recognition capability at the moment
>so that's why I chose a good quality camera. I can probably take the
>casing off to save some weight too. Next year I would like to develop
>a visual gudance capability, which would be really cool and far more
>accurate than GPS.
The ground based autonomous vehicles have the advantage of really only
needing image recognition in 2D. In the air, you have to recognize
and locate objects in 3D, which literally adds a new dimension to the
problem.
I guessed that you were going to be flying this year. If you're just
building parts of the puzzle on the ground this year, there's no
reason to economize on weight and power at this point. Might was well
go with your selected camera.
>I think I have my antenna setup sorted- although I have another
>question: If I use a 2-way splitter on the aircraft to feed the 2
>antennas, will there be any problems with recieving a signal from the
>ground station? What I mean is; is a splitter a bi-directioinal
>component or does it only allow a signal to travel in one direction?
Excellent question and good thinking. A splitter (also known as a
Wilkinson power divider) divides the signal in two. It's totally
passive and bi-directional. At 2.4GHz, it's just two 1/4 wave strip
lines and a 100 ohm resistor. See diagram at:
<http://www.microwaves101.com/encyclopedia/Wilkinson_splitters.cfm#twoport>
<http://www.qsl.net/yu1aw/2G4spliter2.gif>
<http://www.qsl.net/yu1aw/2G4spliterN.gif>
Let's assume that there's no internal losses for now. A 2.4GHz signal
applied to the input/output port, will be divided equally between the
two other ports. That means that each antenna will get half the RF
power or 3dB. 3dB is NOT a large drop in signal level. For example,
the rule of thumb for range versus power is a 6dB drop in power equals
a 50% drop in range. 3dB loss is equal to about 0.7 times the range.
If possible, simply increase the transmit power 3dB (double the
milliwatts) and you'll be exactly the same as before.
The situation is also good in receive. There is no loss between
either antenna port and the input/output port. Every dB received by
either antenna is delivered directly to the input/output port.
There is also considerably isolation (10-30dB) between antenna ports.
This is a huge help in keeping the antenna termination impedance near
the desired 50 ohms, thus keeping the VSWR and losses low. If you can
tolerate minimal isolation, the 100 ohm resistor can be eliminated.
However, there are some potential problems. RF radiating from both
antennas simultaneously will cancel and add at various points. This
will create nulls and peaks in the antenna pattern. The basic idea is
that each antenna should not "see" each other. More simply, there
should be no position, where you can see both antennas at the same
time. That's not going to be easy with the donut shaped pattern of a
common vertical 1/4 wave antenna. The patterns will overlap and
possibly cancel. However, it is possible with cardioid patterns, such
as patch or panel antennas.
There's also the question of where to mount the antennas. On top and
on the belly are the obvious locations. However, I think you'll get
better hemispherical coverage with one antenna mounted in the nose,
pointing forward, and another out the tail pointing aft. This allows
the body to provide some shielding between antennas. There may be
some signal loss dead ahead, which will be critical during landing.
However, that will also be the point of closest approach so distances
will be minimal.
Since you're not flying this year, I suggest you start with just one
antenna. Play with it in a mockup or model and see what it does. Hang
the model from a string in a tree, and measure the received signal
strength. That will determine if two antenna are necessary.
Also, I suggest you consider separating the video, data, and control
links. Control can be done at much lower frequencies (27, 72MHz)
using conventional RC hardware. This has the advantage of being
fairly orientation insensitive. I don't think you have the 900MHz
band in UK, so that's out. 2.4GHz is it for data and video, but I
suspect that the video will saturate the link leaving no time slots
for data. If there's some other frequencies available in UK for
video, I would use it.
Gotta run... good luck.
--
Jeff Liebermann
jeffl@cruzio.com
150 Felker St #D
http://www.LearnByDestroying.com
Santa Cruz CA 95060
http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
Re: webcam : autonomous aircraft project 
Linear Mode
