Is it possible to change the dimensionality of the measurement matrix dynamically with the OpenCV Kalam filter. For example as more features are observed in a scene I want to add these to the Kalman filter to be predicted along with the previously observed features.