Tracked states

After calling the advance method, the tracker updates the states $\lbrace\mathbf{x}\rbrace$. To access the state vectors $\lbrace\mathbf{x}\rbrace$, we have to inspect the Kalman filter objects in the tracks stored in the tracker object.

For example, to access the first track of the first class

...
track = tracker.tracks_c[0][0]

Note, that there will be at least one object class in the tracker, even if the tracker object was instantiated without specification of the maximal number of classes.

Each track possesses one Kalman filter object (OpenCV cv2.KalmanFilter class). Therefore, we can inspect the posterior state as in the snippet below

...
vec_x = track.kkf.kalman_filter.statePost[:, 0]

Note that the internal distribution of variables in the state vectors are dimension-sorted. It is often more convenient to use derivative-sorted states.