简介：ThetraditionalorientedFASTandrotatedBRIEF(ORB)algorithmhasproblemsofinstabilityandrepetitionofkeypointsanditdoesnotpossessscaleinvariance.Inordertodealwiththesedrawbacks,amodifiedORB(MORB)algorithmisproposed.Inordertoimprovetheprecisionofmatchingandtracking,thispaperputsforwardanMOKalgorithmthatfusesMORBandKanade-Lucas-Tomasi(KLT).ByusingKalman,theobject’sstateinthenextframeispredictedinordertoreducethesizeofsearchwindowandimprovethereal-timeperformanceofobjecttracking.TheexperimentalresultsshowthattheMOKalgorithmcanaccuratelytrackobjectswithdeformationorwithbackgroundclutters,exhibitinghigherrobustnessandaccuracyondiversedatasets.Also,theMOKalgorithmhasagoodreal-timeperformancewiththeaverageframeratereaching90.8fps.

简介：Thispaperpresentsahandheld3Dvision-basedscannerforsmallobjectsbyusingKinect.Itisdifferentfromthepreviouscolor-glove-basedapproacheswhichrequiresegmentingthetargetobject.First,weeliminatethenoisesandtheoutlierscausedbyholdinghands.Second,weapplyKinect-fusionalgorithmandtruncatedsigneddistancefunction(TSDF)torepresent3Dsurfaces.Third,weproposeamodifiedintegrationstrategytoeliminatethehandeffect.Fourth,wetakeadvantageoftheparallelcomputationofGPUsforreal-timeoperation.Themajorcontributionsofthispaperare(1)theregistrationprecisionisimproved,(2)theofflineamendmentandloopclosureoperationarenotrequired,and(3)concave3Dobjectreconstructionisfeasible.IndexTermsHandheld3Dscanning,Kinect-fusion,Truncatedsigneddistancefunction(TSDF).1.IntroductionRecently,thesensor-based3Dmodelreconstructionmethodshavebeenproposed[1].Thesensordeviceshavedifferentpropertiessothatthe3Dreconstructionalgorithmsvaryaccordingly.Thecommonlyusedsensordevicesaretime-of-flight(ToF)cameras[2]-[4],laserscanners[5],andstructuredlightscanners[6],[7].Lasershavegainedareputationforaccuracy;however,caremustbetakentouseeye-safelaserswhenoperatinginproximitytohumans.Foraninteractivesystem,thestructuredlightscannerwhichisbasicallyapassivevision-basedsensordeviceissuperiorbecauseitprovidesa2DdepthimageperframeandismoreaccuratethanthatofaToFcamera.Here,wepresentareal-time3DscannerusingthedepthimagescapturedbyKinect.

简介：僵硬声学的散布由的部件的分离在水下目标在获得如此的目标的结构的特征是必要的。克服看起来有一样的僵硬结构的问题光谱在时间领域的结构，时间频率窗帘来源分离(BSS)能与图象形态学在联合被使用分开不同目标的僵硬散布部件。基于一个热点模型，有时间频率分发的目标的僵硬散布结构的分离被推出。用一个词法过滤器，在Wigner-Ville分布(WVD)的不同特征观察到术语能被简化移开任何跨术语的干扰单个汽车术语和十字。由选择术语表明的汽车的时间和频率点，BSS的精确性能被改进。试验性的模拟被使用了，与在播送信号，相对振幅和时间延期参数的脉搏宽度的变化，以便分析这个新方法的可行性。模拟结果证明新方法不仅能分开僵硬散布当有弹性的散布并且僵硬散布时，部件，而且罐头也分开部件同时存在。试验性的结果证实新方法能在分开僵硬散布结构被使用在水下目标。

简介：

简介：