Bernoulli filter for detection and tracking of an extended object in clutter

An extended or non-point object is characterized by multiple feature (or scattering) points which can generate measurements/detections. The problem is joint track formation and maintenance for an extended moving object. Due to imperfect detection, only some of the feature points are detected while false alarms (or clutter) can also be reported. Standard tracking techniques assume point objects, that is at most one detection per object, and hence are not adequate for this problem. The paper presents a theoretical solution in the form of the exact Bayes filter, referred to as the Bernoulli filter for joint detection and tracking of an extended object. The derivation follows the random set filtering framework introduced by Mahler and is based on the binomial random finite set model of object generated measurements. The filter is implemented approximately as a particle filter and its performance analyzed numerically by simulations. An application to video tracking is also presented.