基于残差网络的目标参数预测

打开文本图片集

中图分类号：TP182 文献标志码：A文章编号：1003-3114（2025）04-0832-12

Abstract：Parameter estimationforcone-shapedspacetargetsplaysacrucialroleintaskssuchastargetatitudedetermination， re-entrylandingointpredictio，andtargetrecognition.Intisstudyadeepleaingaproachisploedtopredictgeoicad motionparametersofconshapedspacetargetsundervrioustyesoffalsetargetinteference.Fistlyleveragingteonceptoffacet decompositionfromphysicaloptics，multipletargetfacetscatteringmodelsareestablished，andsimulationsofchosignalsfromvarious targetsareconducted.Subsequently，aShort-TieFourierTransfor（ST）isappdtoanalyeehosignalsintime-frqencydo main，enablingextractionofmicro-Dopplertime-frequencyspectrainducedbytargetmicro-motions.Finaly，basedonheroposed SpatialPyramidPoling-ResidualNetwork（SPP-ResNet），parameterpredictionforcone-sapedpacetargetsisachieved.Siulation results demonstrate that the proposed method achieves an Mean Relative Error（MRE）of less than 15% in parameter predictions，laying afoundation for accurate trajectory prediction of cone-shaped space targets.

Keywords：lidar；micro-Doppler；optical signal processing;deep learning；parameter identification

0 引言

物体的微动会引起激光雷达回波的额外频率调制，从而产生目标多普勒频率的侧带，称为微多普勒效应[1]。（剩余15471字）