Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Implementing the Pedestrian class

The nature of the Kalman filter provides the main rationale for creating a Pedestrian class. The Kalman filter can predict the position of an object based on historical observations and can correct the prediction based on the actual data, but it can only do this for one object. As a consequence, we need one Kalman filter per object tracked.

Each Pedestrian object will act as a holder for a Kalman filter, a color histogram (calculated on the first detection of the object and used as a reference for the subsequent frames), and a tracking window, which will be used by the MeanShift algorithm. Furthermore, each pedestrian has an ID, which we will display so that we can easily distinguish between all of the pedestrians ...