last update 24/03/2009
This page is part of the
"easy-introduction-to-the-Kalman-filter" series that we started
recently on this site. It should be considered as an illustration of what
has been explained so far. The main idea is to apply the simplest scalar
Kalman to a colour tracking program that runs under ROBOLAB/LabVIEW. For
best compatibility we chose ROBOLAB254 version. In order to get an idea of
what the filter is capable in the domain of image processing, please
- For those who are not familiar with the
ROBOLAB approach to graphics, the program diagram has been well
commented and should be self-exlanatory.
- The program only observes and estimates
the x-coordinate. nothing prevents from minimally altering the program
to include the y-coordinate. Since both coordinates are independent, the
K-step function can be used twice in the same iteration then. Also take
notice that the state variable x could represent the
vector (x,y)T in that case.
- The prediction model is based on the
estimation of the object-speed.
- There is only one simple validitycheck of
the measurement. If the BLOB-analyser doesn't find any area, then the
measurement variance is switched to 1000. This obviously isn't
sufficient for a serious detection of the absence/presence of the
tracked object. We could add a certain area threshold and many other
image processing configuration. But this is not relevant here for the
illustration of the use of the Kalman filter.
1. Track x-coordinate
2. Track both
TO KALMAN FILTER INDEX PAGE