Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Initializing and applying the Kalman filter

We looked at some aspects of the Kalman filter's initialization earlier, in the Initializing the tracker section. However, in that section, we noted that some of Kalman filter's matrices need to be initialized or reinitialized multiple times, as the application runs through various frames and various states of tracking or not tracking. Specifically, the following matrices will change:

• The transition matrix: This matrix expresses the temporal relationships among all the output variables. For example, this matrix can model the effects of acceleration on velocity, and of velocity on position. We will reinitialize the transition matrix every frame because the frame rate (and, therefore, the time step ...