简介:Thispaperproposesanovelregistrationmethodforaugmentedreality(AR)systemsbasedonOrientedFASTandRotatedBRIEF(ORB)andFastRetinaKeypoint(FREAK)naturalfeatures.IntheproposedORBFREAKmethod,featureextractionisimplementedbasedonthecombinationofORBandFREAK,andthefeaturepointsarematchedusingHammingdistance.Togetgoodmatchingpoints,cross-checksandleastmediansquaresareusedtoperformoutlierfiltration,andcameraposeisestimatedusingthematchedpoints.Finally,ARisrendered.Experimentsshowthattheproposedmethodimprovesthespeedofregistrationtobeinrealtime;theproposedmethodcanaccuratelyregisterthetargetobjectunderthecircumstancesofpartialocclusionoftheobject;anditalsocanovercometheeffectsofrotation,scalechange,ambientlightanddistance.
简介:一个专业化电气化学的测量房间被插进一个飞行员水分发系统模仿管子内部墙的腐蚀。线性polarisation抵抗(LPR)技术和电气化学的阻抗光谱学(EIS)在实时被测量学习腐蚀率(CR)和生铁的规模的变化。三个腐蚀阶段根据LPR分析被观察:有显著地波动的CR的一个起始的阶段,有慢慢地减少的CR的一个发展阶段,并且有约0.1575公里的低CR的一个稳定的阶段?晡整?潖摮镲??鲹秪琀牥獩楴?畣癲?湡污獹獥?敒?瑬?桔?敲畣牲湥散朠潲灵攠桸扩瑩摥猠杩楮楦慣瑮祬栠杩敨??楤敭?倨??????湡???倨???????敬敶獬愠摮愠琠敲摮琠睯牡?楨桧牥瀠慬整敬?敬敶獬???????桔?灯楴慭?畣潴晦瘠污敵?敷敲ㄠ??e并且衰退emission-to-output比率在动态综合气候经济模型,描述了投射的碳排放是有艾尔2包含锡的O3-MgO-TiOx核心被发现是??
简介:滑动模式控制的适应backstepping与输入浸透为不明确的非线性的系统的一个班被建议。一个命令过滤了途径被用来阻止输入浸透破坏神经网络(NN)的适应能力。控制法律和NN的适应更新法律在Lyapunov功能的意义被导出,因此稳定性能甚至在输入浸透下面被保证。建议控制法律对骚乱是柔韧的,并且它能也消除输入浸透的影响。模拟结果显示建议控制器有好表演。
简介:ThewaterdistributionsystemofoneresidentialdistrictinTianjinistakenasanexampletoanalyzethechangesofwaterquality.Partialleastsquares(PLS)regressionmodel,inwhichtheturbidityandFeareregardedascontrolobjectives,isusedtoestablishthestatisticalmodel.TheexperimentalresultsindicatethatthePLSregressionmodelhasgoodpredictedresultsofwaterqualitycomparedwiththemonitoreddata.Thepercentagesofabsoluterelativeerror(below15%,20%,30%)are44.4%,66.7%,100%(turbidity)and33.3%,44.4%,77.8%(Fe)onthe4thsamplingpoint;77.8%,88.9%,88.9%(turbidity)and44.4%,55.6%,66.7%(Fe)onthe5thsamplingpoint.
简介:Anewmethodforidentifyingnonlineartime-varyingsystemswithunknownstructureispresented.Themethodextendstheapplicationareaofbasissequenceidentification.Theessentialideaistoutilizethelearningandnonlinearapproximatingabilityofneuralnetworkstomodelthenon-linearityofthesystem,characterizetime-varyingdynamicsofthesystembythetime-varyingparametricvectorofthenetwork,thentheparametricvectorofthenetworkisapproximatedbyaweightedsumofknownbasissequences.Becauseofblack-boxmodelingabilityofneuralnetworks,thepresentedmethodcanidentifynonlineartime-varyingsystemswithunknownstructure.Inordertoimprovethereal-timecapabilityofthealgorithm,theneuralnetworkistrainedbyasimplefastlearningalgorithmbasedonlocalleastsquarespresentedbytheauthors.Theeffectivenessandtheperformanceofthemethodaredemonstratedbysomesimulationresults.