Thread Subject: Tracking a point. Image acquisition from a webcam

Hi everybody,

I am trying to develop a matlab program able to track a point with a differentiable characteristic (such as a color). Think for example in a red point moving in a white background. What I want to do is obtain the coordinates of this point respect some reference, and evaluate the rate of the traking (number of positions detected per second) when this point is moving.

I am newbie in this issue. I have read something about the Kanade-Lucas-Tomasi algorimth. Right now I am "studying" the image adcquisition toolbox help.

Is there any program done similar for what i want to do?

Any help will be useful,

Kind regards.

MRR wrote:
> Hi everybody,
>
> I am trying to develop a matlab program able to track a point with a differentiable characteristic (such as a color). Think for example in a red point moving in a white background. What I want to do is obtain the coordinates of this point respect some reference, and evaluate the rate of the traking (number of positions detected per second) when this point is moving.
>
> I am newbie in this issue. I have read something about the Kanade-Lucas-Tomasi algorimth. Right now I am "studying" the image adcquisition toolbox help.
>
> Is there any program done similar for what i want to do?
>
> Any help will be useful,
>
> Kind regards.

I don't think that this demo uses the same tracking algorithm as you
want to use, but it solves a similar problem:

http://www.mathworks.com/products/imaq/demos.html?file=/products/demos/shipping/imaq/demoimaq_Pendulum.html

-Dave

"MRR " <mario_ruz@hotmail.com> wrote in message <ga8e1i$p3r$1@fred.mathworks.com>...
> Hi everybody,
>
> I am trying to develop a matlab program able to track a point with a differentiable characteristic (such as a color). Think for example in a red point moving in a white background. What I want to do is obtain the coordinates of this point respect some reference, and evaluate the rate of the traking (number of positions detected per second) when this point is moving.
>
> I am newbie in this issue. I have read something about the Kanade-Lucas-Tomasi algorimth. Right now I am "studying" the image adcquisition toolbox help.
>
> Is there any program done similar for what i want to do?
>
> Any help will be useful,
>
> Kind regards.

Can you assume continuous motion? Like the particles are passing over top of one another, not bouncing? That makes it a lot harder. I've seen projects where they use Kalman filtering to predict the next location to recover from occlusion.

If not, can't you just take the number of points, and assign them via a least distance from the results of the previous assignment?

~Adam

I Adam

My application can be simple, so I dont consider the case in which particles pass over top of the object I want to track.

Anyway right now I dont know how to figure it out. I am trying to work with OpenCV libraries but right now i am not able of finding a solution for my problem.

regards

"Adam " <not.real@email.com> wrote in message <ga8kr2$n57$1@fred.mathworks.com>...
> "MRR " <mario_ruz@hotmail.com> wrote in message <ga8e1i$p3r$1@fred.mathworks.com>...
> > Hi everybody,
> >
> > I am trying to develop a matlab program able to track a point with a differentiable characteristic (such as a color). Think for example in a red point moving in a white background. What I want to do is obtain the coordinates of this point respect some reference, and evaluate the rate of the traking (number of positions detected per second) when this point is moving.
> >
> > I am newbie in this issue. I have read something about the Kanade-Lucas-Tomasi algorimth. Right now I am "studying" the image adcquisition toolbox help.
> >
> > Is there any program done similar for what i want to do?
> >
> > Any help will be useful,
> >
> > Kind regards.
>
> Can you assume continuous motion? Like the particles are passing over top of one another, not bouncing? That makes it a lot harder. I've seen projects where they use Kalman filtering to predict the next location to recover from occlusion.
>
> If not, can't you just take the number of points, and assign them via a least distance from the results of the previous assignment?
>
> ~Adam
>