Vor dem Regler: Ein Verfahren zur Mehrfach-Objektverfolgung füur Fahrerassistenzsysteme (Before the Controller: A Multiple Target Tracking Routine for Driver Assistance Systems)

Abstract

Abstract This paper presents a probabilistic approach to multiple target tracking with applications to driver assistance systems. The authors develop the probabilistic theory behind two routines that together handle uncertain target motion and uncertain measurement origin. The first algorithm incorporates multiple target motion models and acts like a self-adjusting, variable-bandwidth filter to accurately track a single target during all possible motion. The second algorithm allows for multiple targets to be tracked simultaneously by performing a computationally-bounded, measurement-association process. The complementary nature of these two routines is highlighted and their ability to incorporate measurements from multiple sensors is also discussed. The effectiveness of this probabilistic approach is illustrated through experimental test results.