Multisensor particle filter cloud fusion for multitarget tracking
Daniel Danu, Thia Kirubarajan, Thomas Lang, Michael McDonald · International Conference on Information Fusion · 2008
Within the area of target tracking particle filters are the subject of consistent research and continuous improvement. The purpose of this paper is to present a novel method of fusing the information from multiple particle filters tracking in a multisensor multitarget scenario. Data considered for fusion is under the form of labeled particle clouds, obtained in the simulation from two probability hypothesis density particle filters. Different ways of data association and fusion are presented, depending on the type of particles used (e.g. before resampling, resampled, of equal or of different cardinalities). A simulation is presented at the end, which shows the improvement possible by using more than one particle filter on a given scenario.