(Dr. Nader Sadegh, advisor)
"High Speed Target Tracking Using Kalman Filter and Partial Window Imaging"
A method to increase the overall sampling frequency of a vision sensor for visual servoing applications was developed. The frequency is increased, by reducing the number of pixels that must be acquired and processed. This was done, by creating a predictor that estimates the future target positions. A great deal of information can be computed using vision sensors. Extracting this, information, however, is time consuming. Currently the sample time of vision sensors limits its feasibility in control applications. An algorithm has been designed to track an object, with unknown trajectory, that is moving within the cameraís field of view (FOV). The algorithm acquires and processes only the region of pixels that contain the target. To select the correct region of pixels, a discrete steady state Kalman filter was used. The Kalman filter uses the measured position of the targetís centroid as well as previous state estimates to determine the centroid position in the next time step.