where kalman filter fails in multiobject tracking?

Why KF can't track jitter and creep conditions? can i get code for MTT using Particle filter?

Answer 1

[Sampada]

Yes. i am using the same example. But couldn't change parameters of KF. in fact i dont know how to do it. Even can't see coding of correct() and predict() functions. Please help me in doing that. I am using BBC motion gallery videos for trial. I don't get proper results when used abrupt motion change videos

Answer 2

[Dima Lisin]

The KF is implemented as a class called vision.KalmanFilter. correct() and predict() are its methods. You can see the code by doing

 >> edit vision.KalmanFilter

But you do not need to change the code. The example is using the configureKalmanFilter() function to create the KF object. The function lets you specify the motion model, the initial error covariance, and the noise covariances as diagonal matrices.

If you want additional flexibility in specifying the parameters, you can use the constructor of vision.KalmanFilter directly.

For more information, please see this example on how to set up the Kalman filter .