| Traditional target tracking algorithms assume that the target is a point target,i.e.that a target produces at most one measurement at each moment in time.As sensor accuracy increases,this assumption no longer holds and a target can generate multiple measurements at each moment,with the number of measurements varying over time,making the target tracking problem known as extended target tracking.As the number of measurements increases,one is not only interested in the kinematic state of the target,but also in obtaining an estimate of the target’s shape.Many scholars have studied the extended target tracking problem,but there are still many problems that need to be solved.In this paper,the extended target tracking algorithm based on Gaussian surface fitting is carried out to address the problems such as poor shape estimation and no better evaluation index,and the main work is as follows.1.To address the current problems of poor extended target shape estimation and inability to reflect the main features of the shape,this paper proposes an extended target shape estimation algorithm based on level sets and Gaussian surface fitting.The algorithm uses multiple moments of measurements and Gaussian surface fitting to build a distribution model of the measurement sources containing shape information,and uses level sets to describe the shape of the extended target,on the basis of which the shape estimation of the extended target is performed.To address the problems of modelling the shape function and thresholding in the level set method,a connection between the shape function and the spatial distribution function of the measurement source is established,and an adaptive algorithm for solving the threshold of the level set is proposed.Aiming at the problem of pre-determined parameters in the traditional Gaussian surface fitting method,an adaptive algorithm for Gaussian surface covariance based on edge measurement is proposed,which can effectively solve the problems existing in the traditional algorithm.The proposed shape estimation algorithm is able to obtain a more accurate estimated shape without any shape prior.Simulation experiments show that the proposed algorithm can obtain better estimation results even if the shape of the target is complex.2.The proposed shape estimation algorithm uses multiple momentary measurements to estimate the shape of the target.To address the problem of increased computational effort caused by the increase in the number of measurements,a recursive formulation of the estimated shape is proposed to be obtained using a Gaussian mixture model in a Bayesian framework to update the estimated shape.In estimating the kinematic state of the extended target,a volumetric Kalman filter with a high filtering accuracy is used to estimate the kinematic state of the extended target.The simulation results show that the proposed algorithm can estimate the kinematic state of the target more accurately,the estimated shape can show the shape details and main features of the target,and the Gaussian mixture model can also effectively update the estimated shape of the target and reduce the algorithm computation.3.An extended target tracking algorithm evaluation metric based on a radial function is proposed to address the existing extended target shape evaluation metric that uses a value to measure the size and shape similarity between the estimated shape and the true shape.The metric uses the standard deviation of the radial ratio to describe the degree of similarity between the estimated shape and the true shape,and the mean value of the radial ratio to describe the size relationship between the two shapes.A radial management strategy is also proposed for the problem of the existence of multiple radii under one angle.Simulation experiments show that the proposed metrics are able to better evaluate the similarity between the estimated shape and the true shape,regardless of the differences in detail or size between the estimated shape and the actual shape. |
PDF Full Text Request