how can i track a set of pixels using kalman filter?

to track object from frame to frame, I wanna track a defined pixels that determines the object, using kalman filter... I don't know from where I begin

[Roger Labbe]

You can try my free Kalman filtering book here: https://github.com/rlabbe/Kalman-and-Bayesian-Filters-in-Python

It is in Python, not Matlab, but once you lean the basic ideas you can either write your own code or download one of several libraries available for kalman filtering.

I didn't answer your computer vision part. That's a whole different problem. Ideally you would want a feedback loop. The Kalman filter predicts where to look for the pixel. You run image detection to try to find the pixel. If where you found it matches the prediction, you are doing well. If not, well, that's where the difficulty starts. Generally you will have either 0 detected pixels, or multiple detected pixels. Look into authors such as Lawrence Stone, or Bar-Shalom. The wikipedia article https://en.wikipedia.org/wiki/Radar_tracker gives you some common algorithm names to start investigating.

Despite the tutorials on youtube and such that make this seem trivial this is in general quite a difficult problem. But the above is the journey you need to take, though I skipped the details of image detection, which is its own field.

Answer 1

[Dima Lisin]

There is a vision.KalmanFilter object in the Computer Vision System Toolbox. Take a look at the Motion-based Multiple Object Tracking example.