Adaptive Target Tracking With Interacting Heterogeneous Motion Models

Multiple motion estimators such as an interacting multiple model (IMM) have been utilized to track target objects cars and pedestrians with diverse patterns. However, the standard IMM has limitations in combining models different state definitions, so it cannot contain a complementary set of that accurately work for all In this paper, we propose IMM-based adaptive tracking heterogeneous velocity representations linear/curvilinear models. It can integrate four definitions dimensions be completely complimentary types motions. We experimentally demonstrate effectiveness proposed method accuracy various patterns using two datasets: synthetic datasets real datasets. Experimental results show achieves also cars, pedestrians, drones world.