预览加载中，请您耐心等待几秒...

基于单目相机的空间非合作目标姿态测量(英文).docx

预览
1/2
2/2

在线预览结束，喜欢就下载吧，查找使用更方便

下载提示 文本预览

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

基于单目相机的空间非合作目标姿态测量(英文) Introduction Attitudeestimationofspaceobjectsisanimportanttaskinthefieldofrobotics,spaceexploration,anddefense.Itisusedtodeterminetheorientationofatargetinspace,whichiscrucialinvariousapplicationssuchasdocking,landing,andreconnoitering.Inthispaper,wepresentamethodfornon-cooperativeattitudeestimationofaspaceobjectbasedonamonocularcamera. Background Theattitudeofaspaceobjectcanbedeterminedbymeasuringtherelativeorientationofitsfeatureswithrespecttoareferenceframe.Severalmethodshavebeenproposedforattitudeestimationusingvarioussensorssuchasgyroscopes,accelerometers,andmagnetometers.However,thesesensorshavetheirlimitationssuchasdrift,noise,andinterference.Ontheotherhand,vision-basedapproacheshavegainedsignificantattentionduetotheirhighaccuracy,flexibility,andlowcost. Methodology Inourproposedmethod,weassumethatthespaceobjecthasknowngeometricalfeaturessuchasedges,corners,orpatches.Wealsoassumethatthecameraiscalibratedanditsintrinsicparameterssuchasfocallength,principalpoint,anddistortioncoefficientsareknown.Thealgorithmconsistsofthefollowingsteps: 1.Imageacquisition:Weacquireasingleimageofthespaceobjectusingthemonocularcamera. 2.Featuredetection:WedetectthegeometricalfeaturesofthespaceobjectusingafeaturedetectorsuchasHarriscornerdetector,SIFT,orSURF.Thesefeaturescanberobustlydetectedandmatchedeveninthepresenceoflightingchanges,clutter,orocclusion. 3.Featurematching:Wematchthedetectedfeaturesinthecurrentimagewiththefeaturesinthereferenceimage.Thereferenceimagecanbeeitherapriorimageofthespaceobjectoratemplateimage. 4.Perspective-n-point(PnP)algorithm:WeusethePnPalgorithmtoestimatetheposeofthespaceobjectwithrespecttothecamera.ThePnPalgorithmrequiresatleastfourmatchedfeaturecorrespondencesandassumesapinholecameramodel. 5.Refinement:WerefinetheinitialestimateusinganiterativemethodsuchasLevenberg-Marquardt.Thismethodminimizesthereprojectionerrorbetweentheobservedfeaturesintheimageandtheircorresponding3Dprojectionsontheobject.Thereprojectionerrorisameasureo