webcam tracking project

[SeZioNe13]

very good

[samuraiml]

No I didn't use openCV. I wanted to implement some algorithm myself :) After I've seen the openCV demos though my next attempt is going to involve it for sure :P

[RightQuiteNot]

Is this built on OpenCV?

[debmita10]

Hi: I am on a research project in medical imaging, and we need to track patient motion. Is there anyway you can share your code with us?

Thanks.

[TheYoucef2009]

@debmita10 Hi, i am also on research, i propose active contour tracking if you want the résult is juged good by experts, you can contact me

[samuraiml]

@TheYoucef2009 : Hi, just took a look at the active contour model. Seems like a good fit for refining the result, since I can already guess an initial position for the 'snake' using the method shown here. I'm definitely going to try that :)

[TheYoucef2009]

@samuraiml First thank's for responding,

in my my methode, i used manual or auto initialisation, with circle or ellipse or square shape for the initial contour and the implementation by fast level set witch gives a good result, i used chan&vese and geodesic modele advantages to construct my modele and the result is very interesting :)

[samuraiml]

@debmita10 Sure :) just give me a mail address in a pm or sth :)

[ParkFlyyer]

@AzraelUK1 Yes it can

[TheYoucef2009]

hi

i'am beginner

i want to know which program to use

and if c++ then is it the win32 api

that get pixel and set pxel

[ParkFlyyer]

C++ is ok, but I use java script to make my programs

[TheYoucef2009]

thanks for responding;

so ,is it in real time or recorded video?

[ParkFlyyer]

It can be in both real and recorded

[TheYoucef2009]

i am working with active contour method, i dont know if you can help me so i work in Matlab and in real time

:)

[wilwad]

nice. I'm stuck with mine. What language did you use? I use VB6, yes, it does the job :D. Care to share code?

[samuraiml]

I was trying to achieve that style of interface in the first time :)

1) It does locate many markers at once - the red piece of paper I used has too much specularity, so the algorithm sees it as a lot of blobs (clusters of red pixels among white stripes that represent the specular color). I use a statistical method to reduce the result to one (which is the case for this video) or 'n' clusters, by separating "distant" blobs and averaging "close" ones.

2) I haven't thought about rotaion yet.

[themanyone]

That is similar to the algorithm I used. It can keep track of multiple objects, even of the same color, and it even sorts them by pixel count, giving an approximate measurement of depth. With a little more work it would be able to discern crude geometries.