第36卷第2期2023年2月传感技术学报CHINESEJOUＲNALOFSENSOＲSANDACTUATOＲSVol.36No.2Feb．2023项目来源:国家自然科学基金项目(11961001);宁夏自然科学基金项目(2018AAC03126);宁夏高等学校一流学科建设(数学学科)(NXYLXK2017B09);北方民族大学重大专项项目(ZDZX201801);宁夏智能信息与大数据处理重点实验室开放基金(2019KLBD004)收稿日期:2021－09－18修改日期:2022－05－235-DofCircularFeature’sPoseVariablesEstimationAlgorithmBasedonExtendKalmanFiltering*WUYangyu1*，LICui2(1．LibraryNorthMinzuUniversity，YinchuanNingxia750021，China;2．CollegeofMechanicalEngineering，BeijingInstituteofTechnology，Beijing100081，China)Abstract:Adrawbackoftheexistingcircularfeature’sposevariablesestimationalgorithmsisthatinputframesareprocessedindepend-ently，so，valuabledynamicinformationoftargetisignored，andtheposeestimationaccuratecanstillbeimproved．Ahighaccuracy5-Dofcircularfeature’sposevariablesestimationmethodisproposedbyusingExtendKalmanfiltering，andcircularfeature’sposeisre-presentedbya5-DoF(degreeoffreedom)vectorofξ=(X，Y，Z，α，β)T．Inthismethod，bayesianframeworkisintroducedtocapturethetimeinformationofcontinuousvideoframes，andthecircularposeestimationsystemisoptimized．Firstly，theproposedalgorithmgivesprojectionrepresentationof3Dobjectcircularfeaturetointeractwiththeellipsecontourintheimagebyusingξ，andthenthenon-linearmeasurementfunctionisobtained．Later，themeasurementfunctionandExtendKalmanfilteringarecombinedforposeoptimization．Inaddition，Kalmanfilteringisutilizedtocorrectposeestimationresult，whichprovidesarobustmethodforobject2D－3Dposeestimationbyusingcircularfeatureinachallengingenvironment．Theexperimentalresultsshowthattheprecisionofcircularfeature’sposeestima-tioniseffectivelyimprovedbyutilizingthetimecorrelatedinformationforimagesequencewithnoisesofdifferentsquareerror．Keywords:poseestimation;circularfeature;projectionmodelof3dobjectcircularfeature;extendKalmanfilteringEEACC:7230doi:10．3969/j．issn．1004－1699．2023．02．017基于扩展卡尔曼滤波的5-Dof圆位姿估计算法*吴仰...
