IM3HT algorithm: A joint formulation of IMM and MHT for multi-target tracking

IMM (Interacting Multiple Model) and MHT (Multiple Hypothesis Tracking) are today interesting techniques in the tracking field. Specifically, IMM is a filtering technique where r standard filters cooperate to match the true t arget model; MHT is a multi-scan correlation logic, which defers data assoc iation until more data are available so to reduce the risk of mis-correlati on. The combination of IMM and MHT promises improved tracking performance. This paper provides a theoretical formulation of a joint IMM plus MHT algor ithm, which we shall term IM3HT; also, the paper includes the results of pe rformance comparison with a 'classical' MHT in terms of tracking errors.