Adaptive Processing to Ensure Practical Application of a Multiple Hypothesis Tracking System

摘要

Modern computational capabilities allow the practical application of Multiple Hypothesis Tracking (MHT) for difficult tracking conditions. However, even in typical expected scenarios, periods of unusually high target and / or clutter density may occur that stress the ability of MHT to operate in real-time and under the constraints of limited computer memory. This paper outlines methods that are being developed to ensure practical application, even though some performance degradation must be accepted, during these difficult conditions. These methods include the adaptive choice of track and hypothesis pruning parameters, IMM filtering models and new track initiation strategies as a function of the latency between the time that current observations are received and the track processing time. Methods to ensure that memory constraints are satisfied are also discussed. The methods are illustrated with examples from simulated missile defense scenarios where periods of very high target density are expected and a ground target tracking scenario with real radar data.