Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

MATLAB Video tracking

Not open for further replies.


Newbie level 3
Jun 21, 2011
Reaction score
Trophy points
Activity points
I need help for video tracking in MATLAB using correlation algorithm. I have MATLAB 6.5. Please guide me from where to start. I can read the video in MATLAB. hHow to play the video? How to create a template?

Please give me more details. Is the video captured and stored in a memory or do you want to do it in real time?

also, 'video tracking' is a bit ambiguous. Do you want to track a moving object ?

I would like to do real time tracking of the object. Object is acquired using a CCD camera. Right now I have to develop a code for the project. The object is preferably moving.
Initially template matching is performed using correlation. Then Kalman filter predicts the next position of target. If Kalman filter is mislead then fast means shift algorithm is applied.

correlation is an extremely simple task.Have you sorted out the camera and MATLAB interface?

I don't know about the kalman filtering part. I used sum of absolute differences to track motion.

First you'll have to acquire the image, identify the object using autocorrelation then, convert it into a logical image, then find centroid, then do all processing using it. Is this what you are planning? or something different?

Thank you for your response.............

1.From a CCD camera i'll be acquring a video.
2.Through Frame Grabber,i'll capture the images from that video.
3.I'll read those captured images into MATLAB.
4.I won't find the centroid i'll correlate two frames captured from frame grabber.
5.I also have to set a threshold value through a thresholding function.

So for that I'll need a correlation algorithm and a thresholding algorithm.
Since you don't use Kalman method I'll manage it myself. Basically Kalman filter predicts the next position of the object.
If Kalman filter fails, I'll use a mean shift algorithm.

Not open for further replies.

Part and Inventory Search

Welcome to