Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

C programming for computer vision

2 views
Skip to first unread message

K C L Yong

unread,
Nov 12, 1997, 3:00:00 AM11/12/97
to

I am writing a program in C to read in the grey levels of a
digitised image from a camera such as to detect any moving object and
then tracking that object with the camera mounted on a motor.

Any help will be greatly appreciated. Please e-mail me.

Lionel

Jonathan G Campbell

unread,
Nov 12, 1997, 3:00:00 AM11/12/97
to K C L Yong

K C L Yong wrote:
>
> I am writing a program in C to read in the grey levels of a
> digitised image from a camera such as to detect any moving object and
> then tracking that object with the camera mounted on a motor.
>
You don't mention your timescale, or the level of the project, e.g.
Bachelors final year project, or whatever. Nor are there any details of
the camera/framegrabber.

My major advice is to get your systems engineering done properly. There
are a number of tasks/subsystems which, at least in the beginning, must
be considered separately:

1 Image acquisition. 1.1 Start off by reading one image -- into a file.

1.2 Make sure you can read that image into an array in a C program -- or
wherever you propose to do the processing.

1.3. Acquire a sequence of images -- either moving camera or moving
object.

2. Moving object detection. 2.1 Compare two successive images etc. See
the literature. 2.2 Try something over a sequence of images.

3. Object tracking -- camera following object. You need to set up a
control problem which has some objective, e.g. centroid of object at
centre of image. Define an 'error', e.g. vector displacement of object
from centre. Devise a 'control-law': error -> camera control

Okay, the comments above are short on detail; but, try solving the whole
problem at once and you're lost. Maybe, if you're lucky, right at the
end, you'll be able to integrate your subsystems. If it's for an exam.,
the systems approach is also important -- you want marks for the
individual parts, even if you never get a complete system working.

One point. If you don't already have an software image processing
system, even the first steps will be quite demanding. For example,
trying to debug an object detection algorithm will be horrendously
difficuly unless you have good interactive feedback -- not unlike
operating blindfold. Therefore, see if they have any interactive image
processing software available, e.g. Matlab -- or Khoros, cheap by
universities CHEST licence.

If your professors mean the latter to be part of the project, then be
methodical, e.g. build a suite of programs:

- read from camera and write to some standard file type (one that you
can read into an array);

- various processes, e.g. edge detection, difference (between two
files),

- display.

If you need to develop such software, you may get some hints in:

Jonathan Campbell and Fionn Murtagh
Signal and Image Processing in Java.
Presented at IMVIP '97, University of Ulster, Magee College,
Derry, 10-13 September 1997.

available from: http://www.infm.ulst.ac.uk/research/preprints.html

although that software is in Java, and, generally, a good deal more
ambitious than what you require.

Good luck.

Jon Campbell

--
Jonathan G Campbell Univ. Ulster Magee College Derry BT48 7JL N. Ireland
+44 1504 375367 JG.Ca...@ulst.ac.uk http://www.infm.ulst.ac.uk/~jgc/

blueeyedpop

unread,
Nov 12, 1997, 3:00:00 AM11/12/97
to

sci.image.processing
K C L Yong wrote in message <3469A2...@shef.ac.uk>...

>I am writing a program in C to read in the grey levels of a
>digitised image from a camera such as to detect any moving object and
>then tracking that object with the camera mounted on a motor.
>

Mark C. MacNab

unread,
Nov 13, 1997, 3:00:00 AM11/13/97
to

I bought a Quick Cam 2 for this purpose. I would be interested in any
progress you amke especially regarding the techniques for recognizing an
object ot he shape of an object. Please stay in touch. I won't be able to
get to this for a while because I am creating a set of robot building
blocks. I hope to have voice recognition and vison in the toolset one day.
My goal is not to become filthy rich but to provide the means that we can
start to build really cool robots including household robots quickly. The
idea to create an open archtecture that can be added to by any one. I see
this as the same sort of thing that was happening in the early days of Home
Computers. I hope that robots become the next fad and either replace or
augment the home computer. Robots will become the arms and legs of the
intelligent house in my lifetime (and I am just recovering from a heart
attack).

Sorry. I got carried away.

K C L Yong <ELA9...@shef.ac.uk> wrote in article

G. Grunt

unread,
Nov 13, 1997, 3:00:00 AM11/13/97
to

Have you thought of using a grey scale Connectix Quickcam? It's only
about $100 and can install via the parallel port. If the project is
running under Windows NT there are drivers available which makes it easy
to program; no low-level programming. If you want to use the camera on
other OS systems (DOS, UNIX, etc.), there are actually many 3rd party
drivers. These are much lower-level but you have more control and may be
able to speed up the acquisition rate and fine-tune it for your
particular application. The camera has a pretty decent CCD. Not sure
what resolution you are looking for though. If this sounds like
something you are interested in I can locate the web sites for you.

Hope this helps.

0 new messages