基于主动特征选择的非合作航天器鲁棒视觉导航方法研究.pdf

《基于主动特征选择的非合作航天器鲁棒视觉导航方法研究.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、136上海AEROSPACE航天SHANGHAI第33卷2016年第6期文章编号：1006—1630(2016)06一0136一06基于主动特征选择的非合作航天器鲁棒视觉导航方法研究宁明峰1，张世杰1，张翰墨2(1．哈尔滨工业大学卫星技术研究所，黑龙江哈尔滨150080；2．上海航天控制技术研究所，上海201109)摘要：面向非合作目标航天器近距离操作任务，针对采用自然特征的单目视觉相对住姿参数确定过程中特征提取与匹配导致的粗大误差增加导致结果不准确甚至错误，以及特征数量多增大计算量等问题，提出一种融合随机采样一致性(RANSAC)算法和主动特征选择的鲁棒

2、视觉导航方法。用RANSAC算法剔除有粗大误差的特征点，给出了基于RANSAC的特征点选择步骤；根据不同特征点组合所计算的克拉美罗(CRLB)不同，用参数化CRLB下限选择对位姿确定精度有显著影响的点以减少参与计算的特征数量，给出了基于CRLB的特征点选择流程。仿真结果表明：综合RANSAC和CRLB的特征点选择方法可显著减少特征点数量，提高了位姿解算精度。关键词：非合作目标；视觉导航；特征点选择；RANsAC；CRLB；特征点数；鲁棒性；位姿精度中图分类号：V448．2文献标志码：ADOI：10．19328／j．cnki．1006—1630．2016．0

3、6．020Rob吣tMethodStudyofActiVeFeatureSelectionforNon-CooperatiVeSpacecraftVision-BasedNaVigationNINGMing—fen91，ZHANGShi—jiel，ZHANGHan—m02(1．ResearchcenterofSatelliteTechnology，HarbinInstituteofTechnology，Harbin150080，Heilon西iang，China；2．ShanghaiInstituteofSpaceflightControlTechnolo

4、gy，Shanghai201109，China)Abstract：Tosolvetheproblemthattheerroroffeaturepointsextractingormatchingandthenumberoffeaturepointswouldleadtoinaccurateresultsorthewrongresultsandhugeamountofcalculationontherelativepositionandattitudeparameterdeterminationduringnon—cooperativetargetspace

5、craftproximityoperations，arobustmethodforvisionnavigationfusingtherandomsampleconsensus(RANSAC)algorithmandanactivefeatureselectionmethodwasputforwardinthispaper．FirstthegrosserrorwaseliminatedbyRANSACalgorithm．TheselectionstepsforfeaturepointsweregivenbasedonRANSACalgorithm．Thent

6、hedifferentpoints，whichhadsignificantimpactondeterminingprecisionbasedonCram亡r—Rao10werbound(CRLB)，wereselectedtoreducethenumberoffeatureinvolvedinthecalculationaccordingtothedifferentCRLBcalculatedfromvariousfeaturepointssets．ThefeaturesselectionflowchartwasgivenbasedonCRLB．Thesi

7、mulationresultsshowedtheselectedpointsbycombiningofRANSACandCRLBcouldbereducedandtheprecisionofpositionandattitudehadbeenimproVed．Keywords：Non．cooperativespacecraft；Visualnavigation；Featurepointselection；Randomsampleconsensus(RANSAC)；Cram∈r-Rao10werbound(CRLB)；Featurepointnumber；R

8、obust；Precisionofpositionandattit