Box Particle Filter’s Application To Extended Target Tracking

Because of the great potential in military and civilian areas and the key technologies in modern air-defense war and the field of aerospace, target tracking has always attracted much attention from the international and domestic academics. With the research and further development of broadband radar, its higher range resolution and strong anti-jamming performance make wideband radar play an increasingly important role in electronic warfare, and the issue arose under the case of "high range resolution" that the extended target tracking is also urgent to be resolved. In extended target tracking, more than one measurements are produce per time, so during multiple extended targets tracking, complex data association results often make them difficult to be tracked and the algorithm also has higher computational complexity. How to reduce the computational complexity and the impact on target state estimate caused by immovable clutter measurement in partition cells, as well as a mandatory requirement to the measure generated by the target must obey certain distributions and so on. How to ensure a well tracking performance under these cases has become a challenging research.In this thesis, under the condition of interval measurements, the author mainly investigates box particle filter‘s application in extended target tracking.1. The basics of the box particle filter are studied, and through existing research to make sure the scope of its application and implementation process, using the interval analysis methods to deal with non-precision measurements. Box particle filter is the combination of interval analysis and particle filter, replacing the accurate data with the interval data. Box particle filter with tens of box particles can get the similar tracking performance as the traditional filter with thousands of particles. It has the advantages of low computational complexity and a better performance in the case of data with fuzziness.2. Multiple extended targets tracking with Interval Analysis is studied. First use the box particle to deal with the extended range of the target, and use a rectangular area to replace the extended target for tracking. The paper presents a novel PHD Filter for Tracking Multiple Extended Targets Using Interval Analysis. It can reach a well performance with a low computational complexity while reducing the requirement of the measurements’ distribution. Due to the special likelihood function that don’t need to distinguish the source of measurements in the division unit, it can eliminate the impact of clutter measurement within the division unit to get a better performance and also can directly track the extended targets with different sizes.3. The maneuvering extend target tracking algorithm with box particle is researched. Based on using Interval Analysis to deal with extend target and the interacting multiple model, this paper presents an IMM filter for tracting maneuvering extend target using box particle. The new born box particles can reduce the impact of the wrong model estimation. Then combine with the RFS to present an interacting multiple model box particle PHD filter(IMM-BPF-PHD) algorithm and apply it on multiple extended targets tracking. The proposed algorithm inherits the advantages of box particle filter on extend target with a small number of particles, fast runtime, strong anti-noise performance and the measurements without the need to obey a certain distribution.