 |
BorlandTalk.com
Borland discussion newsgroups
|
| View previous topic :: View next topic |
| Author |
Message |
Bob Bedford
Guest
|
 Posted: Mon Jun 14, 2004 2:54 pm Post subject: Re: Movement recognition |
 |
|
| Quote: | wow, it's really cool idea...
An old idea...but other people already tought about it.. and implement it as |
well.
| Quote: | Detect a dot from webcam is not very difficult, but I think there are many
issues are much more difficult than motion detect... |
The problem is that this operation shouldn't take too much CPU and also be
very fast using the cams at 30fps (for smoothing movements). I don't want
motion detection....many companies and universities are working hard on
it...I only want to get the movement of a colored dot. I'm talking about
glasses, but this colored dot may be placed also in the mouth, on the front
or over a cap (the best idea is to put it on the microphone headset if
any...) Anyway, if it's so simple to detect a dot being of a certain color
out of a webcam signal, could you please tell me how ??? I've even never get
any signal from a webcam on a Delphi program, so I don't know where to
start.
| Quote: | The Glasses position change may not equal to user's sign-direction change.
User move his body or seating position will also produce a glasses
movement... |
In any case, you can set the center of the dot position at any time by
pressing a certain keyboard button. There is also to set an amount of "dead
zone" where the dot has no effect. The other idea is that more the head is
turned, quicked is the movement. After moving his body, the user may press
the button and the center of the PAN is set again.
| Quote: | Anyway, you can set some constraints for your application. Like asking the
user don't move during using the glasses..  |
Gamers rarely change body position. They are always face to the screen so
the head would normally be on the same place during the game.
images coming from a webcam...
Have a look at those sites:
http://www.ews.uiuc.edu/~awu/pix/earthlight.html
http://www.mousevision.com/game/index.htm
the first one is exactly what I'm looking for, but with only one dot, as I
don't need 3dimensional move, I only want left-right and top-bottom, as the
second site I send you does.
|
|
| Back to top |
|
 |
RandomAccess
Guest
|
| Posted: Mon Jun 14, 2004 3:02 pm Post subject: Re: Movement recognition |
|
|
Hi Bob,
I don't see how this can be usefull for a game if you are actually using a
monitor - unless the monitor moves too. If you move your head, you will no
longer be looking directly at the center of the screen - also there are then
limitations on far you can actually move your head and still see the screen
at all.
So, it would be impossible for example to do a full 360 degree turn using
thhis technique - unless you are talking about using a head movement to
start an "automatic" turn which can then be stopped by anohter head
movement.
Is this what you had in mind?
best regards
|
|
| Back to top |
|
|
Bob Bedford
Guest
|
| Posted: Mon Jun 14, 2004 3:37 pm Post subject: Re: Movement recognition |
|
|
| Quote: | I don't see how this can be usefull for a game if you are actually using a
monitor - unless the monitor moves too. If you move your head, you will
no
longer be looking directly at the center of the screen - also there are
then
limitations on far you can actually move your head and still see the
screen
at all.
So, it would be impossible for example to do a full 360 degree turn using
thhis technique - unless you are talking about using a head movement to
start an "automatic" turn which can then be stopped by anohter head
movement.
Is this what you had in mind?
|
Yes,
If you have a joystick and use then POV hat of it in a virtual view (in
Flight simulator 2002, for example), you move your head aroud by pushing the
button on any direction. Once you don't push the button, the head stop
moving. This is good, but once you have moved your button following an other
aircraft (in flight combat simulator), you don't know in wich direction you
are looking.... The idea is to turn you head 15° in any direction. If you
are at the 15° left, then you are looking on far left of your vision. If you
look say at 7° left, you look at middle possible left. I mean If you can PAN
to 180° on the left, moving your head 15°, you screen show you the 180°, but
you can still look at the screen, and at the same time, you know where you
are looking (on the left). This is particularly interesting when you do a
looping and you want to quickly see at the left or right, or the top, to see
if you are doing your looping well aligned or are changing direction, or
when you follow an aircraft in a close combat and you don't really know
where you are looking with your POV. This should also allow you to move
quicker than with the POV hat. This will be useful in a car simulation game:
you can look left and right in an easier way. You turn a little your head,
but a lot in the game, and the monitor is still on his place.
In any case, you still look at you screen, but your head tell the game where
to turn the view, and how much. The limitation of the head movement is how
far you can turn your eyes to still see what's on the screen. This amount of
head turn gives the maximum of the PAN turn.
Have a look at http://www.edimensional.com/products/trackir3.htm
There is a little video showing how it works
http://media.naturalpoint.com/video/trackIR_small.wmv
This device is really too expensive, and I know that can be done with a
normal webcam.
|
|
| Back to top |
|
|
RandomAccess
Guest
|
| Posted: Mon Jun 14, 2004 6:21 pm Post subject: Re: Movement recognition |
|
|
Hi Bob,
| Quote: |
In any case, you still look at you screen, but your head tell the game
where
to turn the view, and how much. The limitation of the head movement is how
far you can turn your eyes to still see what's on the screen. This amount
of
head turn gives the maximum of the PAN turn.
Have a look at http://www.edimensional.com/products/trackir3.htm
There is a little video showing how it works
http://media.naturalpoint.com/video/trackIR_small.wmv
This device is really too expensive, and I know that can be done with a
normal webcam.
|
Interesting. I am still not convinced this is a good idea. Ideally, you
want to detect the minimum amount of movement that is "practical". I simply
can't see a game being playable when you need to look out of the corner of
your eye - I guess it would take some getting used to.
Detection of movement
=====================
This starts to get really complicated because the scene can "move" for a
variety of reasons - other than just turning your head - you must remember
that the only data you have is a video frame and in my opinion that's simply
not enough in order to produce a good workable system. The 2 dots reference
points are ok for detecting initial movement, but you need to "know" how
that movement relates to the scene. A lateral movement isn't the same as
twisting your head but would be probably detected the same.
Actually, when I think about it, I'd prefer to move my head slightly to
each side rather than twist it. At least that way, I always look directly
at the screen.
I just can't see this system working without additional reference data being
available in the video frames - if you want only head-twisting.
best regards
|
|
| Back to top |
|
|
Samson Fu
Guest
|
| Posted: Tue Jun 15, 2004 1:06 am Post subject: Re: Movement recognition |
|
|
You can use Directshow to capture video from camera, DSpack is a good component set for doing that. You can take a look at
SampleGrabber sample.
http://www.progdigy.com/dspack/
--
Samson Fu
"Bob Bedford" <bedford1 (AT) YouKnowWhatToDoHerehotmail (DOT) com> wrote
| Quote: | wow, it's really cool idea...
An old idea...but other people already tought about it.. and implement it as
well.
Detect a dot from webcam is not very difficult, but I think there are many
issues are much more difficult than motion detect...
The problem is that this operation shouldn't take too much CPU and also be
very fast using the cams at 30fps (for smoothing movements). I don't want
motion detection....many companies and universities are working hard on
it...I only want to get the movement of a colored dot. I'm talking about
glasses, but this colored dot may be placed also in the mouth, on the front
or over a cap (the best idea is to put it on the microphone headset if
any...) Anyway, if it's so simple to detect a dot being of a certain color
out of a webcam signal, could you please tell me how ??? I've even never get
any signal from a webcam on a Delphi program, so I don't know where to
start.
The Glasses position change may not equal to user's sign-direction change.
User move his body or seating position will also produce a glasses
movement...
In any case, you can set the center of the dot position at any time by
pressing a certain keyboard button. There is also to set an amount of "dead
zone" where the dot has no effect. The other idea is that more the head is
turned, quicked is the movement. After moving his body, the user may press
the button and the center of the PAN is set again.
Anyway, you can set some constraints for your application. Like asking the
user don't move during using the glasses..
Gamers rarely change body position. They are always face to the screen so
the head would normally be on the same place during the game.
You can find more information about color spacing from elf's website:
http://www.efg2.com/Lab/Library/Delphi/Graphics/Color.htm
Thanks for the link, but the first problem is how to get and decode the
images coming from a webcam...
Have a look at those sites:
http://www.ews.uiuc.edu/~awu/pix/earthlight.html
http://www.mousevision.com/game/index.htm
the first one is exactly what I'm looking for, but with only one dot, as I
don't need 3dimensional move, I only want left-right and top-bottom, as the
second site I send you does.
|
|
|
| Back to top |
|
|
Jacques Oberto
Guest
|
| Posted: Wed Jun 16, 2004 9:11 am Post subject: Re: Movement recognition |
|
|
Hi Bob,
I have just uploaded the code to "borland.public.attachments"
Cheers,
Jacques
"Bob Bedford" <bedford1 (AT) YouKnowWhatToDoHerehotmail (DOT) com> a écrit dans le
message de news:40d0045f (AT) newsgroups (DOT) borland.com...
| Quote: | Thanks Jacques,
But the downlaod link doesn't work.
It would have been perfect.
|
|
|
| Back to top |
|
|
Bob Bedford
Guest
|
| Posted: Thu Jun 17, 2004 2:28 pm Post subject: Re: Movement recognition |
|
|
Great, thanks for the code !!! I couldn't have the code since the author's
site is prohibited and he didn't respond at my email.
Thank you again.
Cheers
Bob
PS: you are french ??? I've seen your reply is frech...T'es d'où ??? je suis
en Suisse...
"Jacques Oberto" <jo (AT) nospam (DOT) com> a écrit dans le message de
news:40d00ecd$1 (AT) newsgroups (DOT) borland.com...
| Quote: | Hi Bob,
I have just uploaded the code to "borland.public.attachments"
Cheers,
Jacques
"Bob Bedford" <bedford1 (AT) YouKnowWhatToDoHerehotmail (DOT) com> a écrit dans le
message de news:40d0045f (AT) newsgroups (DOT) borland.com...
Thanks Jacques,
But the downlaod link doesn't work.
It would have been perfect.
|
|
|
| Back to top |
|
|
Jacques Oberto
Guest
|
| Posted: Thu Jun 17, 2004 3:54 pm Post subject: Re: Movement recognition |
|
|
You're welcome.
J.
P.S. Je suis à Paris
"Bob Bedford" <bedford1 (AT) YouKnowWhatToDoHerehotmail (DOT) com> a écrit dans le
message de news:40d1aa7c$1 (AT) newsgroups (DOT) borland.com...
| Quote: | Great, thanks for the code !!! I couldn't have the code since the author's
site is prohibited and he didn't respond at my email.
Thank you again.
Cheers
Bob
PS: you are french ??? I've seen your reply is frech...T'es d'où ??? je
suis
en Suisse...
|
|
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
