The challenges of the multiple target tracking problem include the uncertainty in the number of objects and the uncertainty in the track to measurement association. These uncertainties are due to object appearance and disappearance as well as missed detections and false alarms. A rigorous approach to model the multiple target tracking problem is the multi-target Bayes filter using random set methods. The Probability Hypothesis Density (PHD) filter approximates the multi-target Bayes filter using the first statistical moment in order to obtain a computationally tractable algorithm. Due to increasing sensor resolutions, each target tends to generate multiple measurements. Consequently, the commonly used assumption that a target generates at most one measurement is not valid any more and a representation as extended targets is required.
Within this talk, the multiple target tracking problem and the PHD filter are briefly introduced. Further, I will give examples of tracking of extended targets and group targets using the PHD filter and show results using both video data and laser range data.
Biography:
Karl Granström received the MSc degree in Applied Physics and Electrical Engineering in June 2008, and the PhD degree in Automatic Control in November 2012, both from Linköping University, Sweden. He works as a post doctoral fellow at the Division of Automatic Control at the Department of Electrical Engineering at Linköping University. His research interests include sensor fusion, tracking of extended targets, and random set methods for multiple target tracking.