A Four-Model Based IMM Algorithm for Real-Time Visual Tracking of High-Speed Maneuvering Targets

Abstract

In recent years, visual tracking algorithms based on state estimators have been developed in order to improve the performance during tracking tasks. However, this performance changes according to target type, object kinematics and scenario complexity. When working with high-speed maneuvering targets, tracking errors increase considerably due to low response of estimators as well as the kinematic mismatch betwen the real motion profile and the one assumed by the estimator. Some examples of objects that present this high-speed behavior are rockets, aircrafts and missiles. To overcome this visual tracking problem, this work proposes an interacting multiple model algorithm based on four kinematic models: constant velocity, constant acceleration, constant turn and thrust acceleration. We present three different scenarios with complex maneuvers for comparison study, and experimental results show that visual tracking is improved when using the proposed strategy.