Re: Re: camera recommendation motion tracking – active IR markers

Sep 30, 2007 at 4:15am

Re: Re: camera recommendation motion tracking – active IR markers

that’s a pretty fast FPS for the price range….

have you used the camera in a max/msp/jitter environment before?

the unibrain stuff works remarkably well

I am also currently working on an IR tracking system – what are you using for your markers?

best
jamie

—– Original Message —-
From: Phil Donahue
Sent: Saturday, September 29, 2007 10:28:56 PM
Subject: [jitter] Re: camera recommendation motion tracking – active IR markers

Thanks very much for your reply.

I have used a track IR camera from Naturalpoint before and noticed how accurately and smoothly it can track for determining mouse cursor position:

http://www.naturalpoint.com/trackir/02-products/trackir-product-comparison.html

These cameras are supposed to have fps values of >100 (or ‘sample rate’ – same thing?). Is this a critical factor for smooth motion tracking? Or is resolution more important? I am worried that the unibrain running at 30 fps may not be smooth enough.

#33901
Sep 30, 2007 at 4:48am

Phil, i am also interested in the naturalpoint camera. if you can give more
information on connecting it to jitter that’ll be great.
naturalPoint those some heavy duty processes in hardware, that is what nice
about it. a basic chain of cv effects like.
grab>barrel correction>denoising>background removal>tracking can get
expensive with the cpu.
i moved most of it to the gpu (check the archive) but it real depend on what
exactly you want to track. but a basic rule is that for tracking dots you
need less resolution and more fps.
i think the naturalpoint is a good choice.

On 9/29/07, jamie jewett wrote:
>
> that’s a pretty fast FPS for the price range….
>
> have you used the camera in a max/msp/jitter environment before?
>
> the unibrain stuff works remarkably well
>
> I am also currently working on an IR tracking system – what are you using
> for your markers?
>
> best
> jamie
>
> —– Original Message —-
> From: Phil Donahue
> Sent: Saturday, September 29, 2007 10:28:56 PM
> Subject: [jitter] Re: camera recommendation motion tracking – active IR
> markers
>
>
> Thanks very much for your reply.
>
> I have used a track IR camera from Naturalpoint before and noticed how
> accurately and smoothly it can track for determining mouse cursor position:
>
>
> http://www.naturalpoint.com/trackir/02-products/trackir-product-comparison.html
>
> These cameras are supposed to have fps values of >100 (or ‘sample rate’ -
> same thing?). Is this a critical factor for smooth motion tracking? Or is
> resolution more important? I am worried that the unibrain running at 30 fps
> may not be smooth enough.
>
>
>
>
>

#113567
Sep 30, 2007 at 8:53am

Its interesting that you say more fps and less resolution is important for dot tracking. Maybe the unibrain camera is not what I am looking for.

With regards to naturalpoint, unfortunately I don’t know anything about connecting to jitter – thats why I am looking for an alternative (eg the Fire-i unibrain camera board).

Optitrack cameras (also from naturalpoint) come with an API written as a series of com automation interfaces:

http://www.naturalpoint.com/optitrack/support/manuals/OptiTrackAPI_1.1.032.doc

so I guess it must be possible to port image data to Max, but how this would be done I don’t know.

#113568
Sep 30, 2007 at 9:03am

Hi Jamie, I have not used the naturalpoint or unibrain cameras max/msp/jitter environment before.

When you say fast for the price, are you referring to the 30 fps unibrain camera?

For active markers I am using arrays of 850nm LEDs, very similar to the ones you can find on this page:

http://www.naturalpoint.com/optitrack/store/accessories-tracking.html

#113569
Sep 30, 2007 at 10:45am

i am overgeneralizing, you can track a dot using a 160×120 matrix, but it
realy depends on the size of the dot, type of lens and quality of light
if its active ir dots (the naturalpoint are not active, they are reflective,
the camera itself has some ir lights shining the reflective coating on the
dots) then you can get by with less resolution but anyways good lighting
setup is vital.

On 9/30/07, Phil Donahue wrote:
>
>
> Hi Jamie, I have not used the naturalpoint or unibrain cameras
> max/msp/jitter environment before.
>
> When you say fast for the price, are you referring to the 30 fps unibrain
> camera?
>
> For active markers I am using arrays of 850nm LEDs, very similar to the
> ones you can find on this page:
>
> http://www.naturalpoint.com/optitrack/store/accessories-tracking.html
>

#113570
Sep 30, 2007 at 12:00pm

Hi Yair, given that resolution does not have to be so high what do you think is a minimum fps for smooth motion tracking using IR LEDs for markers? Can lower fps values be compensated for (eg with larger markers)? I think you are right about the quality of the lens and the lighting conditions being crucial. You are right that naturalpoint predominantly use reflective markers, but they do also employ active markers (eg the ‘track clip’ for the track IR and the LEDs they have for sale with their optitrack gear).

#113571
Sep 30, 2007 at 9:38pm

Hi all,

I think the fps refers to the sample rate of the gesture you are
tracking, so if your fps is low then your reaction to fast and small
gesures will be slow, in a 15fps about 66ms. if it is 120fps about
8ms, etc.

So if you are doing live things, you want to have the fastest reaction
possible, for which you need to be somewhere between 10-30ms of total
latency( that is including the latency of the capture and any further
processing you’ll do (i.e sound) until you get the final output).

So an additional question for me is always to find how much time does
it take me to capture the image. I have found that capturing firewire
can lead to latencies up to 100ms!!

So how much latency do you get capturing the unibrain or naturalpoint?
what does this latency depend on? compressed vs uncompressed firewire?

Am i on the right track or have I lost the way somewhere??

That is: Latency before processing is

capture latency + resolution (fps) latency??

best,

J

ps. excuse the amount of questions, i am trying to solve some of these
obscure misteries myself

On 9/30/07, Phil Donahue wrote:
>
> Hi Yair, given that resolution does not have to be so high what do you think is a minimum fps for smooth motion tracking using IR LEDs for markers? Can lower fps values be compensated for (eg with larger markers)? I think you are right about the quality of the lens and the lighting conditions being crucial. You are right that naturalpoint predominantly use reflective markers, but they do also employ active markers (eg the ‘track clip’ for the track IR and the LEDs they have for sale with their optitrack gear).
>


Jaime E Oliver LR

joliverl@ucsd.edu
http://www.realidadvisual.org/jaimeoliver
www-crca.ucsd.edu/
http://www.realidadvisual.org

9168 Regents Rd. Apt. G
La Jolla, CA 92037
USA

#113572
Oct 11, 2007 at 2:27am

You can also use a line object after you get your tracking values, then turn the sample rate down. This will free up a lot of resources, and for most gesture or motion tracking it’s fine, and often makes the result smoother anyways. Plus you can tweak the line values, so if they’re at 0 ms, it’s as though it’s not there, for a smoothness comparison. You won’t need nearly as high an FPS in this case. Of course, you’ll be introducing interpolated / non-actual values, but these are often just fine for a lot of applications, and take a heck of a lot less processing to generate.

–CJ

#113573

You must be logged in to reply to this topic.