Variable dimensionality Kalman filter

asked 2015-02-04 10:17:11 -0600

johnla
1 ●1

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.

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2015-02-04 11:47:04 -0600

FooBar
2938 ●9 ●37 http://magazino.eu/

It's possible but your covariance matrix will get huge and the inversions will kill your running time. There are some approaches (e.g. fastslam (2.0) from Thrun) that try to deal with that problem.