Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Understanding the predict and update phases

From the preceding description, we gather that the Kalman filter's algorithm has two phases:

• Predict: In the first phase, the Kalman filter uses the covariance calculated up to the current point in time to estimate the object's new position.
• Update: In the second phase, the Kalman filter records the object's position and adjusts the covariance for the next cycle of calculations.

The update phase is – in OpenCV's terms – a correction. Thus, OpenCV provides a cv2.KalmanFilter class with the following methods:

predict([, control]) -> retvalcorrect(measurement) -> retval

For the purpose of smoothly tracking objects, we will call the predict method to estimate the position of an object, and then use ...