Subject:
|
Re: A Generic Idea
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Sun, 5 Jan 2003 21:17:24 GMT
|
Original-From:
|
Steve Baker <sjbaker1@airmailSPAMLESS.net>
|
Viewed:
|
742 times
|
| |
|
|
Henrik Erlandsson wrote:
> Well, to find outline of objects you'd need a low-res (easy to simplify and
> automatically filters out irrelevant details), color (to avoid two objects
> of the same material getting their outlines fused together by the image
> processing software). A 64x64 image that is automatically reduced to 256
> colors (by a filter like in Photoshop) would be perfect! That would take 4KB
> per image, uncompressed. With simple RLE compression probably much less than
> half that.
Yeah - but the amount of CPU horsepower to do that (on the robot remember)
is comparable to the horsepower you'd need to extract the edges and do some
simple image recognition. So why transmit the image at all?
> The problem is finding a camera that does that BEFORE transmitting the
> picture. The only equipment that is reasonably cheap that does this is the
> cel phones I was talking about.
Yes - but they don't compress images like that. They do adaptive motion
compression. For example - if you are looking into the camera talking
to the person at the other end of the line, the camera notices that your
head isn't moving much and sends just the pixels in the region of your
mouth. You notice this immediately on using one of these things because
any sudden motion in the scene causes the frame rate to drop precipitously.
That's OK for the kinds of things cell phone video is used for - but for
navigating a robot around a room, it's hopeless since the robot is moving
around a lot - the image changes in ways that are really bad for the
compression algorithm.
> Another way to do it would be to append a
> small microcontroller gadget to the camera.
Yes - that's what I'd recommend.
> Either way it means hacking
> someone's image format, but at least with the cel phones the compression is
> already done, and the other end, which has a fast CPU, has all the time in
> the world decompressing it.
But getting the compressed image out of the cell-phone system is an undocumented
back door hack that's likely to be well neigh impossible to pull off. Did you
know that some modern cell phones have a MILLION lines of software code in them?
OTOH, web-cams are designed to be interfaced to computers and all of the
protocols are very well known and documented. The interfaces are standard
ones and are not weird or complicated to decode.
> Hmm I think we're straying from the topic. The generic idea was to provide a
> constant link between robot and a better brain. And the celphone (which is
> prolific in Europe, every teenager has one, and 1/3 of the grown-ups have
> one ;)) seemed to be the most generic of the links. It also seemed simple
> and cheap to use, as compared with a PDA+camera which may run you $500-$800.
Yes - that's certainly true - but I think you are missing the point that
cell-phones are not in any way 'hackable'. There are no good interfaces
for getting at their internal goodies.
> In Sweden, a phonecall and $1 can give you a cel phone, provided you
> subscribe to a certain operator for 24-48 months of course. (Those sneaky
> so-and-so :P) This may not be the case in every country of course.
Well, it's at least that easy in the USA. My cell-phone battery was dying
(it wouldn't keep a charge anymore) and for $24 and a one-year signup contract,
we got three new phones, three email accounts, and enough 'free minutes' to
last us six months!
> Whatever the "better brain" is, we need a decent speed interface from it to
> an RCX. The PDA+USB tower (on the robot) suddenly looks more and more
> attractive, especially if you could tweak BrickOS or sth to make it faster
> (I'm assuming the RCX's IO chips is the bottleneck).
The RCX's CPU is pretty slow by modern standards. But if you are using
a PDA with it, that's irrelevent. The RCX would only have to turn motors
on and off and report back sensor data...that's a *tiny* amount of work.
There is plenty of bandwidth to send motor data and sensor reports back
and forth at insanely high speeds. Nothing on the RCX really needs to be
updated more than (say) 10 times a second - and with only three sensors and
three motors, that's at most 30 bytes per second in each direction.
> But I haven't opened the RCX, maybe there's a port which can be used for
> direct serial, (parallel) or USB cable transfer ? Has anyone built a cable
> data interface PC<->RCX?
Well, the IR interface is basically a serial port - so you probably could
hack the hardware to connect the RCX directly to the PC - but if the PC
is the brains, you don't need much bandwidth to the RCX and the IR link
should be just fine for the job.
If the 'brain' is a PDA, then getting it to talk to the RCX using IR
would make a lot of sense. However, not all PDA's can do that because
the IR port on the RCX is like a TV remote - and not like the irDA port
that most PDA's have. You need a PDA that can act like a 'learning remote'
for your TV.
---------------------------- Steve Baker -------------------------
HomeEmail: <sjbaker1@airmail.net> WorkEmail: <sjbaker@link.com>
HomePage : http://www.sjbaker.org
Projects : http://plib.sf.net http://tuxaqfh.sf.net
http://tuxkart.sf.net http://prettypoly.sf.net
|
|
Message has 1 Reply:
 | | Re: A Generic Idea
|
| (...) Sigh. I wish you would read the rest of my message and discover that I address that issue(!), instead of chopping the messages up into neat pieces which you can "deal with". (...) See? (...) Who talked about live video? I was talking about (...) (21 years ago, 6-Jan-03, to lugnet.robotics)
|
Message is in Reply To:
| | Re: A Generic Idea
|
| (...) Well, to find outline of objects you'd need a low-res (easy to simplify and automatically filters out irrelevant details), color (to avoid two objects of the same material getting their outlines fused together by the image processing (...) (21 years ago, 5-Jan-03, to lugnet.robotics)
|
38 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
