Take a look at the result of Kalman filter tracking. Why is it wrong?

When I ran the file kalman.m in from this tutorial

http://www.mathworks.com/matlabcentral/fileexchange/14243-2d-target-tracking-using-kalman-filter

Here are the result

http://www.flickr.com/photos/58590696@N07/5375929588/ http://www.flickr.com/photos/58590696@N07/5375929552/ http://www.flickr.com/photos/58590696@N07/5375329235/

in which the green circular shows the real position of the the ball (before apply Kalman filter) and the red circular shows the estimate position of the ball (after apply Kalman filter). The result shows that using Kalman filter is worse than not using 开发者_如何学PythonKalman filter.

Why?