Kevin Shea Adams

Hi there, I'm looking to track a small point on a performer's instrument in a low lit space. I simply want to have the x/y values for this point. I've thought about tracking color but imagine the low light might make this difficult. So I've been more inclined to use infrared tracking, using a small infrared reflector and bathing the stage in infrared light, or just placing a small infrared emitting LED on the performer. Either solution would be fine, I am just concerned with cost and mostly simplicity - I need to get the tracking system up and running asap. Any thoughts would be much appreciated. Thanks for any help!

2d-position-tracking-in-low-light

IR leds powered by a 3v watch battery like those led throwies.

ir leds are out there don't get the ones for remote but the high brightness

*but* (most common, halogen) theatre lights also give infrared, and will

null your efforts. consider and experiment.

On Tue, Apr 1, 2008 at 6:05 AM, kevin wrote:

> Hi there, I'm looking to track a small point on a performer's instrument

> in a low lit space. I simply want to have the x/y values for this point.

> I've thought about tracking color but imagine the low light might make this

> difficult. So I've been more inclined to use infrared tracking, using a

> small infrared reflector and bathing the stage in infrared light, or just

> placing a small infrared emitting LED on the performer. Either solution

> would be fine, I am just concerned with cost and mostly simplicity - I need

> to get the tracking system up and running asap. Any thoughts would be much

i would go (went) with active trackers.
IR leds powered by a 3v watch battery like those led throwies.
ir leds are out there don't get the ones for remote but the high brightness
ones. its cheap and it work.
 *but* (most common, halogen) theatre lights also give infrared, and will
null your efforts. consider and experiment.
neon is IRfree, but ugly.

On Tue, Apr 1, 2008 at 6:05 AM, kevin  wrote:

>
> Hi there, I'm looking to track a small point on a performer's instrument
> in a low lit space. I simply want to have the x/y values for this point.
> I've thought about tracking color but imagine the low light might make this
> difficult. So I've been more inclined to use infrared tracking, using a
> small infrared reflector and bathing the stage in infrared light, or just
> placing a small infrared emitting LED on the performer. Either solution
> would be fine, I am just concerned with cost and mostly simplicity - I need
> to get the tracking system up and running asap. Any thoughts would be much
> appreciated. Thanks for any help!
>
> -Kevin
>

I'm thinking to just simply try getting some IR lEDs and filtering out visible light using the developed color film trick mentioned often in other posts.

Can anyone recommend some really strong IR LEDs that would be good for tracking at a distance in low light?

Also, is there any real advantage to using reflective tape and bathing the stage in IR light over simply using an IR LED on the performer? I don't imagine much other conflicting IR sources at the performance, there will just be the projection screen and the performer visible in almost total darkness. The only other light will come from whatever we project which I imagine is not going to interfere much in the IR range.

Any other advice? I am thinking cv.jit.track should do the job nicely enough.

Can anyone recommend some really strong IR LEDs that would be good for tracking at a distance in low light? 

Any other advice? I am thinking cv.jit.track should do the job nicely enough.


On Tue, Apr 1, 2008 at 11:59 PM, kevin wrote:

> I'm thinking to just simply try getting some IR lEDs and filtering out

> visible light using the developed color film trick mentioned often in other

this works, even just as a "gel" on top of the lens. but under (between the

> Can anyone recommend some really strong IR LEDs that would be good for

irdevelopment.com among many, or ebay. problem with leds is most are

directional. so they give most amount of light at specific angle (and narrow

at that), a way to solve this is by scratching the plastic surface of the

led to defuse the light. or putting it inside a pingpong ball.

> Also, is there any real advantage to using reflective tape and bathing the

> stage in IR light over simply using an IR LED on the performer?

finding a wide angle ir only source of light will be harder then using

active trackers. most are fairly narrow or expensive. but i havnt gone

> I don't imagine much other conflicting IR sources at the performance,

> there will just be the projection screen and the performer visible in almost

> total darkness. The only other light will come from whatever we project

> which I imagine is not going to interfere much in the IR range.

projectors are good as they dont go into the infra range at all.

prepare a masking patch to draw out a garbage matte for other shiny stuff in

your tracking area. stuff that will work at home will fail elsewhere, be

flexible and better test the performance location beforehand, in same

> Any other advice? I am thinking cv.jit.track should do the job nicely

goodluck. here is a picture of baton i moded to be tracked. i embedded a

cheap crystal at the led's end to spread the light better.

http://img.photobucket.com/albums/v315/pixelpusher/baton_01-1.jpg

On Tue, Apr 1, 2008 at 11:59 PM, kevin  wrote:

>
> thanks yair,
>
> I'm thinking to just simply try getting some IR lEDs and filtering out
> visible light using the developed color film trick mentioned often in other
> posts.

this works, even just as a "gel" on top of the lens. but under (between the
lens abd the sensor) is better.

>
>
> Can anyone recommend some really strong IR LEDs that would be good for
> tracking at a distance in low light?

irdevelopment.com among many, or ebay. problem with leds is most are
directional. so they give most amount of light at specific angle (and narrow
at that), a way to solve this is by scratching the plastic surface of the
led to defuse the light. or putting it inside a pingpong ball.

>
>
> Also, is there any real advantage to using reflective tape and bathing the
> stage in IR light over simply using an IR LED on the performer?

finding a wide angle ir only source of light will be harder then using
active trackers. most are fairly narrow or expensive. but i havnt gone
there.

> I don't imagine much other conflicting IR sources at the performance,
> there will just be the projection screen and the performer visible in almost
> total darkness. The only other light will come from whatever we project
> which I imagine is not going to interfere much in the IR range.
>

projectors are good as they dont go into the infra range at all.
prepare a masking patch to draw out a garbage matte for other shiny stuff in
your tracking area. stuff that will work at home will fail elsewhere, be
flexible and better test the performance location beforehand, in same
conditions.

>
> Any other advice? I am thinking cv.jit.track should do the job nicely
> enough.

goodluck. here is a picture of baton i moded to be tracked. i embedded a
cheap crystal at the led's end to spread the light better.
http://img.photobucket.com/albums/v315/pixelpusher/baton_01-1.jpg

Thanks again yair for your detailed and helpful response. I've so far built a small array of LEDs and attached them to the performer and done some tests with jitter.

cv.jit.track does a pretty good job tracking the LED, but still I wonder if it would work better in a situation where the stage was bathed in IR light from a large IR source and the performer was wearing reflective tape. I wonder this becuase thie cv.jit.track optical flow algorithm seems to track SHAPES, and a single point in the dark from a few leds forms more of a POINT than a recognizable shape.

I recall your warning about CCTV type IR LED arrays often used for security and how they can have a too narrow angle. But then I saw this:

http://www.columbussupply.com/products/?productid=669&price=1278

Not cheap but claims a 360 degree IR light visible from long distance ($120).

My main problem though has been getting the LEDs bright enough to really be visible from a distance when mounted on the performer. I've tried a ping pong ball, but it was too opaque for my weak IR LEDs to illuminate.

http://www.mouser.com/Search/ProductDetail.aspx?qs=vs%252bWWTB4QKYLfZ6uf%2fu%252b0Q%3d%3d

They have a beam angle of 50 degrees (not bad) and a wavelength of 940nm. If anyone can recommend a brighter LED, or any techniques to improve their output it would be appreciated. Right now I am using 4 of them with a nine volt battery stepped down with a 150 ohm resistor. Not very bright at all.

Thanks again yair for your detailed and helpful response. I've so far built a small array of LEDs and attached them to the performer and done some tests with jitter. 

cv.jit.track does a pretty good job tracking the LED, but still I wonder if it would work better in a situation where the stage was bathed in IR light from a large IR source and the performer was wearing reflective tape. I wonder this becuase thie cv.jit.track optical flow algorithm seems to track SHAPES, and a single point in the dark from a few leds forms more of a POINT than a recognizable shape. 

Not cheap but claims a 360 degree IR light visible from long distance ($120). 

On Sun, Apr 27, 2008 at 9:24 PM, kevin wrote:

> Thanks again yair for your detailed and helpful response. I've so far

> built a small array of LEDs and attached them to the performer and done some

> cv.jit.track does a pretty good job tracking the LED, but still I wonder

> if it would work better in a situation where the stage was bathed in IR

> light from a large IR source and the performer was wearing reflective tape.

> I wonder this becuase thie cv.jit.track optical flow algorithm seems to

> track SHAPES, and a single point in the dark from a few leds forms more of a

> I recall your warning about CCTV type IR LED arrays often used for

> security and how they can have a too narrow angle. But then I saw this:

> Not cheap but claims a 360 degree IR light visible from long distance

if you shine a led (even a fairly cheap one, *straight into* the sensor you

will see it from a long distance, this is not true for reflective

illumination (flooding) , there are wide(er) beam ir leds

> My main problem though has been getting the LEDs bright enough to really

> be visible from a distance when mounted on the performer. I've tried a ping

> pong ball, but it was too opaque for my weak IR LEDs to illuminate.

> They have a beam angle of 50 degrees (not bad) and a wavelength of 940nm.

> If anyone can recommend a brighter LED, or any techniques to improve their

> output it would be appreciated. Right now I am using 4 of them with a nine

> volt battery stepped down with a 150 ohm resistor. Not very bright at all.

the led you use peak at 940nm which is too high, check the sensitivity of

your camera sensor. most are more comfortable with 750-880nm wavelength.

also it claims to have a blue lens filter, am not an optical genius but

looks to me you want a clear white lens on your led.

and one good led is enough for a each pingpong ball.

On Sun, Apr 27, 2008 at 9:24 PM, kevin  wrote:

>
> Thanks again yair for your detailed and helpful response. I've so far
> built a small array of LEDs and attached them to the performer and done some
> tests with jitter.
>
> cv.jit.track does a pretty good job tracking the LED, but still I wonder
> if it would work better in a situation where the stage was bathed in IR
> light from a large IR source and the performer was wearing reflective tape.
> I wonder this becuase thie cv.jit.track optical flow algorithm seems to
> track SHAPES, and a single point in the dark from a few leds forms more of a
> POINT than a recognizable shape.

>
> I recall your warning about CCTV type IR LED arrays often used for
> security and how they can have a too narrow angle. But then I saw this:
>
> http://www.columbussupply.com/products/?productid=669&price=1278
>
> Not cheap but claims a 360 degree IR light visible from long distance
> ($120).

if you shine a led (even a fairly cheap one, *straight into* the sensor you
will see it from a long distance, this is not true for reflective
illumination (flooding) , there are wide(er) beam ir leds

>
> My main problem though has been getting the LEDs bright enough to really
> be visible from a distance when mounted on the performer. I've tried a ping
> pong ball, but it was too opaque for my weak IR LEDs to illuminate.
>
> I am using these IR LEDs
>
>
> http://www.mouser.com/Search/ProductDetail.aspx?qs=vs%252bWWTB4QKYLfZ6uf%2fu%252b0Q%3d%3d
>
> They have a beam angle of 50 degrees (not bad) and a wavelength of 940nm.
> If anyone can recommend a brighter LED, or any techniques to improve their
> output it would be appreciated. Right now I am using 4 of them with a nine
> volt battery stepped down with a 150 ohm resistor. Not very bright at all.
>

the led you use peak at 940nm which is too high, check the sensitivity of
your camera sensor. most are more comfortable with 750-880nm wavelength.
940-950 is for remote control.
what camerair filter are you using?
also it claims to have a blue lens filter, am not an optical genius but
looks to me you want a clear white lens on your led.
and one good led is enough for a each pingpong ball.

http://search.ebay.com/infrared-led-pcs
http://search.ebay.com/infrared-led

thanks again! yes using IR leds that are at the right wavelength is critical, and it seems to me that there is a lot of misinfomation online about this. 940nm are super common mainly I think because they have no visible red glow that the lower freq ones sometimes do and are common in remote controls etc. I read somewhere that the wii sensor bar uses LEDs at 940 but this information is sketchy. 850 is standard for illumination purposes as yair has pointed out.

the camera i'm using is a sony trv-900 (possibly a 950) that actually has a thing called "nd filter" which is i believe a neutral density filter. our performance is in near dark conditions so the only real conflicting light i have to worry about is off the projector. the nd filter seemed to cut it down nicely but i may have to try another method to cut visible light as i mentioned in my earlier posts

as for the cv.jit.label and cv.jit.blobs.sort I'd be curious to see how you're tracking the position of your led point using these objects. i've sort of just reached to cv.jit.track because it works so easily right out of the box so to speak.

On Tue, Apr 29, 2008 at 4:49 AM, kevin wrote:

> thanks again! yes using IR leds that are at the right wavelength is

> critical, and it seems to me that there is a lot of misinfomation online

> about this. 940nm are super common mainly I think because they have no

> visible red glow that the lower freq ones sometimes do and are common in

> remote controls etc. I read somewhere that the wii sensor bar uses LEDs at

> 940 but this information is sketchy. 850 is standard for illumination

ye, if you dont have the equipment to test you have to do a lot of

eyeballing. the specs are wrong, the google is wrong, i am wrong. trust only

camera sensors, optics and filters can be tailord to specific needs.

nintendo can afford to cut thier glass and spec sensors to use only a very

> the camera i'm using is a sony trv-900 (possibly a 950) that actually has

> a thing called "nd filter" which is i believe a neutral density filter. our

> performance is in near dark conditions so the only real conflicting light i

> have to worry about is off the projector. the nd filter seemed to cut it

> down nicely but i may have to try another method to cut visible light as i

sony camers have that nightshot mode but i dont use it as it cuts to low, i

that page says they go to 1000nm but that deosnt show were it peaks, and he

modded his camera to use a diffrent pass filter. dont go there.

better get a semi-pro firwire camera like the unibrain one, its cheap and a

filter can be easly applied. google shows how to make a good ir-pass filter

from old photo negative, it also tought me how to mod a usb camera to taht

effect, but after going firewire i wont go back.

> as for the cv.jit.label and cv.jit.blobs.sort I'd be curious to see how

> you're tracking the position of your led point using these objects. i've

> sort of just reached to cv.jit.track because it works so easily right out of

its all in the help file of cv.jit.sort , cv objects are your best friends.

at least skim the overview and get a feel for them all.

On Tue, Apr 29, 2008 at 4:49 AM, kevin  wrote:

>
> thanks again! yes using IR leds that are at the right wavelength is
> critical, and it seems to me that there is a lot of misinfomation online
> about this. 940nm are super common mainly I think because they have no
> visible red glow that the lower freq ones sometimes do and are common in
> remote controls etc. I read somewhere that the wii sensor bar uses LEDs at
> 940 but this information is sketchy. 850 is standard for illumination
> purposes as yair has pointed out.

ye, if you dont have the equipment to test you have to do a lot of
eyeballing. the specs are wrong, the google is wrong, i am wrong. trust only
yourself :)
camera sensors, optics and filters can be tailord to specific needs.
nintendo can afford to cut thier glass and spec sensors to use only a very
nerrow bandwidth. we cant.

>
> the camera i'm using is a sony trv-900 (possibly a 950) that actually has
> a thing called "nd filter" which is i believe a neutral density filter. our
> performance is in near dark conditions so the only real conflicting light i
> have to worry about is off the projector. the nd filter seemed to cut it
> down nicely but i may have to try another method to cut visible light as i
> mentioned in my earlier posts
>

sony camers have that nightshot mode but i dont use it as it cuts to low, i
belive.
http://www.bealecorner.com/trv900/d8/ir/
that page says they go to 1000nm but that deosnt show were it peaks, and he
modded his camera to use a diffrent pass filter. dont go there.

better get a semi-pro firwire camera like the unibrain one, its cheap and a
filter can be easly applied. google shows how to make a good ir-pass filter
from old photo negative, it also tought me how to mod a usb camera to taht
effect, but after going firewire i wont go back.

>
> as for the cv.jit.label and cv.jit.blobs.sort I'd be curious to see how
> you're tracking the position of your led point using these objects. i've
> sort of just reached to cv.jit.track because it works so easily right out of
> the box so to speak.
>

its all in the help file of cv.jit.sort , cv objects are your best friends.
at least skim the overview and get a feel for them  all.

2d Position Tracking in Low Light