Subject:
|
Re: Directly connecting a digital camera module to the RCX
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Sat, 16 Apr 2005 09:33:35 GMT
|
Viewed:
|
3013 times
|
| |
|
|
In lugnet.robotics, Steve Baker <sjbaker1@airmail.net> wrote:
> Claude Baumann wrote:
> > We started a new ambitious project. And this time, we are not certain that we
> > will succeed. So, any help is welcome.
> >
> > http://www.convict.lu/Jeunes/RoboticsIntro.htm
> > http://www.convict.lu/Jeunes/RCXCam/RCXCam_Journal.htm
> >
> > Philo, Stef, Ralph ... this is a call for help. We want to implement a simple
> > I2C bus to have at least 50kHz that we need to access our camera module.
>
> You could probably succeed in doing that - but why? The RCX doesn't have
> enough memory to load even a single photo and it's teeny-tiny 8 bit CPU
> will take an ETERNITY to process whatever it can get at.
>
> Unless you have a very strange application in mind, you're going to need a
> better computer than the RCX.
>
> ---------------------------- Steve Baker -------------------------
Good reply.
1. The C3088 module allows framing, so we can reduce the data.
2. Why? Because we are in an educational environment, where the bulk is to
teach how to program micro-controllers with the background thought that if the
students learn how to operate the RCX/H8 with all its limitations, they are
disciplined in finding solutions under the conditions of limitations. So, that's
our pedagogical reflection. Those who have been at our web-site, know that we
already have proven what can reasonably be done with the RCX based on these
settings. The goal is to teach them how to address and operate a camera. If they
did this once in a simplified and primitive environment represented by the RCX,
they will know HOW this works and they will be well prepared for "real" computer
applications later in their studies or professional surroundings. Our point of
view is that things are easy, if you take solutions from the shelf and draw a
budget list with all the costly articles you need. Here we ask them : "That's
the stuff, you won't get anything else. Try to get that drop out of the citron."
3. Eternity? Depends on the data-size, the algorithms and the operating system.
If you look at our binaural sensor (under Ultimate Robolab pages) that is made
only by analog parts. We operate RCX sensor sampling at 36kHz for two channels
-together with data-storing in an array- and then a simplified cross-correlation
which takes about 2sec. for 2kB. Not famous, but try to find a similar solved
project at this low cost dimension. Because: the sensor works and we can build a
robot that turns its head towards you, if you are whistling a tune. No one would
have thought that this was possible with an 8-bit processor. When we started at
that time, people told us that you'd need at least an FPGA or 75Mips.
Our intention is to have sufficiently reduced picture data and only use very
fast algorithms. An optical mouse uses reduced "camera" data to deduce the
direction of motion...
Let's see.
This is challenging, and honestly I'm not sure that this time we will succeed.
But, it is such an interesting project.
By the way, I have worked out a real-time DFT algorithm that actually operates
the DFT for any data-size up to 500Hz with the RCX. This is not published yet
and works under Ultimate ROBOLAB. There are several tricks and it may appear
strange that it should be independant of the data size, but this is one of the
results of the algorithm.
|
|
Message is in Reply To:
6 Messages in This Thread:
   
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
