Research And Implementation Of Multiple Extended Targets Tracking Algorithms

Multi-target tracking technology is a research focus and difficulty in information fusion.Because of its high application value both in military and civil areas,multi-target tracking technology is the great attention of scholars both at home and abroad.With the development of modern sensor and the constant improvement of the resolution,an extended target generate multiple measurements in a sampling instant.Extended target tracking technology become a more and more concerned area in the direction of the target tracking.This dissertation mainly involves researches about the multiple extended targets tracking algorithms with particle filter.Firstly,implementation of particle filter and the multi-target tracking theory based on random finite set(RFS)is introduced.The theoretical foundations of the multi-target tracking algorithm based on probability hypothesis density(PHD)filter are introduced in detail,and its particle implementation of point target and extended target is given.This part provides theoretical support for subsequent chapters.Secondly,in order to deal with the problem of the particle degeneracy in multiple extended targets tracking,the improved method of particle filter——particle flow for nonlinear filter(PPF)is introduced to multiple extended targets tracking.PPF is used instead of resampling,which can reduce the computation and improve the efficiency of the algorithm.Based on the simulation results,it shows the new algorithm can deal with particle degradation,but the new method has some shortcomings.Thirdly,aiming at solving the unknown clutter intensity and unknown newly born target intensity problem in extended targets tracking,a novel approach of multiple extended targets tracking based on intensity filter(i Filter)is proposed.The augmented state space is applied to the RFS extended target tracking model to represent the motion and mutual transformation between extended targets and the clutter.The estimation of the clutter model,the number of extended targets and the estimation of the extended targets state are obtained by measuring space.The particle implementation of this algorithm is given,and the simulations show that the algorithm is effective in multiple extended target tracking.In the end,currently the traditional multiple extended target tracking algorithms based on RFS cause a large amount of computing,misclassification and low tracking performance.Therefore,this dissertation proposes an unknown newly born multiple extended targets tracking algorithm based on mean shift iteration.Firstly,the state of newly born targets are obtained by associating the clustered measurements.Then,the centroids of measurement set are obtained by mean shift iteration which maps the multiple measurements to a single measurement.Finally,the proposed algorithm is performed by using the particle filter method.The simulation shows that the proposed algorithm has significantly lower complexity,the higher tracking efficiency and the more robust performance than the traditional algorithm at targets crossing.Our studies show that this property holds not only in simulations but also in real world applications.