基于联合似然函数的多扩展目标广义标签多伯努利滤波器刘艺多姬红兵*张永权(西安电子科技大学西安710077)摘要:高分辨率雷达监视系统可观测到区域内不同形状的多个扩展目标,可靠的形状估计有利于提高扩展目标跟踪性能,并可作为战场态势评估的重要依据。该文针对不同形状多扩展目标跟踪问题,提出一种基于联合似然函数的广义标签多伯努利(JL-GLMB)滤波器,可实现目标数目、航迹以及形状的精确估计。首先,将目标形状建模为星凸集,并利用非线性量测变换滤波器更新GLMB分布中的高斯分量,有效提高扩展目标状态估计精度。然后,通过对数加权融合策略,构造联合似然函数,综合衡量扩展目标和量测单元之间的相似程度。最后,基于吉布斯采样,提出快速计算扩展目标状态后验概率密度的方法,有效提高数据关联的准确率和计算效率。仿真实验结果表明,所提滤波器能够有效估计不同形状的多扩展目标状态,且在杂波环境下具有稳定的势估计。关键词:多扩展目标跟踪;随机有限集;星凸集模型;非线性估计中图分类号:TN953文献标识码:A文章编号:1009-5896(2023)04-1303-10DOI:10.11999/JEIT220213AMultipleExtendedTargetGeneralizedLabeledMulti-BernoulliFilterBasedonJointLikelihoodFunctionLIUYiduoJIHongbingZHANGYongquan(SchoolofElectronicEngineering,XidianUniversity,Xi’an710077,China)Abstract:High-resolutionradarsystemsmonitormultipleextendedtargetswithdifferentshapesinasurveillancearea.Reliableshapesestimationcaneffectivelyimprovetrackingperformanceandarecrucialtobattle-fieldsituationevaluations.Inthispaper,aJointLikelihoodbasedGeneralizedLabeledMulti-Bernoulli(JL-GLMB)filterisproposedtoestimateaccuratelythenumberoftargets,targettracks,andtargetshapes.Firstly,theextendedtargetismodeledasastar-convexset,andGaussiancomponentsintheGLMBdensityareupdatedbythemeasurementtransformationfiltertoimprovetheaccuracyofstateestimation.Then,ajointlikelihoodfunctionisconstructedbylog-weightedfusionstrategytomeasurecomprehensivelythesimilaritybetweenextendedtargetandmeasurementcell.Finally,afastapproximationmethodforposteriorprobabilitydensityisproposedbasedonGibbssampling,whichimprovestheaccuracyandefficiencyofthedataassociation.Simulationresultsshowthattheproposedalgorith...