Heterogeneous sensors fusion target tracking with a hybrid adaptive filter

An image-based maneuver detector can improve the performance of maneuvering target tracking with a radar and an infrared image sensor. But the variable dimension filter (VDF) algorithm has the inherent defaults of hard switching method. Furthermore, a single, motion model is not sufficient for tracking maneuvering targets. The interacting multiple model (IMM) algorithm can avoid these defaults because of model interaction and soft switching, but its performance is no good during non-maneuvering periods. We developed a hybrid adaptive filtering method. This new algorithm combines the advantages of the two algorithms mentioned above. When the maneuver detector declares an onset of maneuver, this technique uses an IMM estimator instead of a single model filter as used in the VDF algorithm. The three algorithms were all tested in the simulation experiments. The performance of the new algorithm is obviously better than that of the IMM algorithm during nonmaneuvering periods and provides better estimations than the VDF algorithm during maneuvering periods especially right after the end of the maneuver. The whole performance of the proposed method is better than IMM and VDF.