A Branching Algorithm for Discriminating and Tracking Multiple Objects.

A recursive branching algorithm for a large class of multiple-object discrimination and tracking problems is described. The algorithm is an approximation of the optimal unbounded tree-branching solution provided by optimal adaptive filtering theory. The algorithm consists of a bank of parallel filters of the Kalman form, each of which estimates a trajectory associated with a certain selected measurement sequence. The measurement sequences processed by the algorithm are restricted to a tractable number of combining similar trajectory estimates, by passing over unlikely measurement-state associations, and by dropping unlikely trajectory estimates. This branch editing is accomplished by various threshold tests based on the innovations sequence and state estimate of each filter. The numerical experiments performed using the algorithm demonstrate the feasibility of tracking several objects against a noisy background. Two background noise density parameters and two ratio parameters analogous to the classical signal-to-noise ratio are defined. Author