OpenCV 4 with Python Blueprints - Second Edition

Using a box tracker with the Kalman filter

First, we have to figure out how to model each car's state. It might be better to start with the observation model; that is, what can we measure about each car?

Well, the object detectors give us boxes, but the way they are presented is not the best physical interpretation; similar to the train example given previously, we want variables we can reason about and that are closer to the underlying dynamics of the traffic. So, we use the following observation model:

Here, u and v are the horizontal and vertical pixel locations of the center of the target, and s and r represent the scale (area) and the ...