Research On Maneuvering Extended Target Tracking Algorithm Based On Random Finite Set

Multi target tracking refers to the use of a series of measurement data with time-varying and clutter interference to estimate the number and state of targets.In the past,when long-distance target tracking or sensor accuracy was low,the tracked target was usually simplified as a point target.However,in the case of close range target tracking and high sensor accuracy,a single target may generate multiple measurement data,such as short-range aircraft,sea vessels,etc.At this point,the point target model is no longer applicable and must be modeled as an extended target.Due to the increase in measurement data,the relationship between extended targets and measurement data is no longer one-to-one but one-to-many,which leads to a very complex data association between sensor reception measurement and extended target state.In 2009,Mahler modeled both the extended target state and measurement data as a Random Finite Set(RFS),avoiding the complex problem of multiple extended target data association.Subsequently,extended target filters based on RFS theory were proposed one after another.However,there are still many problems and challenges in this theory,such as model mismatch during maneuvering target tracking,difficulty in intuitively and accurately estimating the shape of extended targets,and unknown prior information of new targets.Based on the RFS extended target tracking theory,this article delves into the maneuvering extended target tracking method and achieves the following three results:1.In response to the problem that filtering algorithms for single motion models are difficult to adapt to the dynamic changes of maneuvering targets’ motion states,the Interactive Multiple Model(IMM)and parameterized models based on Orientation and Half Axes Lengths of an Ellipse(OAL)are combined into the Extended Targets Probability Hypothesis Density(ET-PHD)filter,The IMM-OAL-PHD filtering algorithm was proposed.This algorithm utilizes the explicit measurement equation constructed by multiplicative noise to better consider the spatial information of the target.In addition,it not only avoids data association issues,but also combines the advantages of the IMM algorithm’s adaptability to different maneuvering models,as well as the high accuracy and small computational complexity of ET-PHD filtering.The simulation experimental results show that the IMM-OAL-PHD filtering algorithm can quickly track multiple maneuvering extended targets with dynamic changes in target motion state.2.In response to issues such as difficulty in intuitively and accurately estimating the shape of extended targets in cluttered environments,and unknown prior information of new targets,a parameterized multi extended target tracking study based on OAL was conducted on the basis of the Extended Targets Cardinality Balance Multi target Multi Bernoulli(ET-CBMe Mber)filter.Firstly,the algorithm utilizes multiplicative noise to construct an explicit measurement equation that considers target spatial information,and derives a closed form solution for the Gaussian mixture implementation of the OAL-CBMe MBer algorithm;Secondly,based on explicit measurement equations,a priori information of the new target is adaptively constructed using known measurement data,taking into account the centroid position and shape state.The simulation experimental results show that this algorithm improves the estimation accuracy of the number and state of extended targets,and can effectively track multiple extended targets.3.Due to the fact that intuitively and accurately estimating the extended target shape can improve target tracking performance and the possibility of model mismatch during maneuvering target tracking,the IMM-OAL-CBMe Mber filtering algorithm is proposed by combining the parameterized model of OAL and IMM into the ET-CBMe Mber filter.The explicit measurement equation constructed with multiplicative noise can better consider the spatial information of the target by linking the measurement with the centroid state and shape state of the target.In addition,the IMM algorithm has self adaptability to different maneuvering models;The ET-CBMe Mber filter transfers the multi Bernoulli RFS parameters,thereby improving the reliability and efficiency of multi target states.The experimental results show that the proposed IMM-OAL-CBMber algorithm can efficiently track multiple extended targets and accurately estimate the number and status of targets.