Research On Extended Targets With Irregular Shape Tracking Algorithms In Complex Environment

With the development of radar technology and sensor technology,high resolution radar is becoming more and more common.When a target occupies more than one radar resolution unit and produces more than one measurement,the target is called an extended target.In the extended target tracking problem,people no longer only pay attention to the position information of the target,but also hope to get the shape state of the target.Starconvex modeling method uses a finite number of radius functions to describe the object shape,and is suitable for modeling irregular shapes.The Gaussian Process can be used to learn the radius function by parameterizing the radius function in the spatial domain to get intuitionistic result.Gaussian process modeling extended target is the focus of this paper.In the extended target tracking problem under high resolution radar,a target needs to be associated with several measurements.Probabilistic Multi-Hypothesis Tracker(PMHT)algorithm is a multi-target tracking algorithm based on maximum-expectation(EM)algorithm.PMHT algorithm associates the measurement with each target based on probability,and the correlation process of each measurement is a mutually independent "soft" decision model,which avoids the characteristic that the computational complexity of traditional data association algorithm increases exponentially with the number of targets.PMHT algorithm is very suitable for extended target tracking problem.Therefore,this paper focuses on the irregular shape extended target tracking algorithm based on GP model and PMHT algorithm.Its main contents are as follows.1.The main modeling methods of extended target,including random matrix model,random hyperplane model,GP model,etc.,are introduced.The GP model is emphasized.A joint state model is established based on the motion state and the extended state of the target,and both of them are estimated simultaneously.Measurement models can be divided into contour model and surface model according to different measurement sources.2.The basic principle of PMHT algorithm and how to introduce Poisson's ratio to feedback track quality are introduced.PMHT algorithm allows multiple measurements to associate a target,and the measurement correlation process is independent of each other,with low computational complexity and easy extension,so it is very suitable for extended target tracking.3.The contour measurement model is used to model the situation where the measurement is distributed along the boundary of the target.In a 2D scenario,this means that the measurement is a discrete point around a curve.When using lidar as a sensor,the profile measurement model is very suitable.This paper introduces the modeling and application of classical GP algorithm in contour model4.The extended tracking problem becomes more complicated when the received measurement information comes from inside the target.The traditional GP method uses random variables as scaling parameters to adapt to surface measurements.However,this method is not suitable for use in PMHT framework,which leads to computational difficulties.In addition,the traditional GP algorithm often uses a fixed number of radii and uniform Angle spacing,which is not applicable in some scenarios.In addition,deformation and splitting may occur in the process of target movement.In order to deal with this problem,ET-GP-S-PMHT algorithm is proposed in this paper.The algorithm combines GP model with PMHT algorithm and is applied in the surface model scene.At the same time,a dynamic adjustment strategy is given to make the algorithm have certain dynamic characteristics and achieve good performance in the dynamic extended target tracking.Finally,the simulation results show that the algorithm is effective.