资源描述:
《基于光流的微小型飞行器室内避障方法的研究》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、ABSTRACTABSTRACTTheabilityofautonomousobstacleavoidanceiswithhighvalueinindoorenvironmentofMAV,butthefullyautonomousobstacleavoidancetechnologyinindoorwithcomplexsceneshasnotbeenachieved.However,thetinyinsectscanuseopticalflowtoforage,takeoff,land,avoidobstacleandflyslotsandsoon,inspiredbyins
2、ects,obstacleavoidancebasedonopticalflowgraduallydrawattentionfromresearchers,anditwithguidingsignificancefortheimprovementofMAV’sintelligencelevel.Currently,themethodofobstacleavoidancebasedonopticalflowislimited.Obstacleavoidanceisachievedbybalancingtheopticalflowfromleftandrightorupanddown
3、,andWallFollowingiscompletedbymaintainingconstantfloworconstantdistancewithoneside.However,thesewaysofobstacleavoidancewillleadtofrequentchangesindirectionincomplexscenarios,andtherotationopticalflowhasagreatinfluenceonthesubsequentopticalflowanalysis,whichcannotguaranteetheaccuracyofobstacle
4、avoidance.Therefore,inordertooptimizethemethodofobstacleavoidanceofMAV,threeaspectsarestudiedasfollows:1.Arobustopticalflowalgorithmbasedonfive-pointrestraintisproposed.Byoptimizingtheselectionmodeofconfidencepoint,opticalflowiscalculatedbyusingfive-pointconstraintopticalflowcalculatingmethod
5、withthecombinationofglobalandlocal.Barron’sopticalflowtestindicatorsareextendedtoassesstheperformance.Experimentsshowthatthismethodcanimprovetheaccuracyandreal-timeperformanceofopticalflowcomputation.2.AmethodofobstacleavoidanceinindoorenvironmentofMAVwithconstantaltitudeisproposed.Thismethod
6、demandstodividethefieldofviewofMAVtoseveralcolumns.Eachcolumnhaveavaluetodenotetheheadingtendency,whichcomesfromopticalflowinformation,TTCinformationandtargetdeviationinformationofthecorrespondingcolumn.ThecolumnswhichhavethehighestvaluearetheheadingdirectionandtheTTCisusedtodecidewhenandhowt
7、oavoidtheobstacle.Thisobstacleavoidanceapproachcannotonlydecreasetheinfluenceofnoisepix,butalsolightenthedependenceontextureofenvironment,andthelowfrequencyofvelocityvarietywilldecreasetheinfluenceofrotationpartofopticalflow.IIABSTRACT3.Resea