简介:针对目前国际上中子评价核数据库中没有n+12C反应在20MeV以上的第一、第二激发态非弹散射角分布数据的情况,利用最小二乘法给出了数据库中缺失的Legendre系数,并且能很好地符合现有的实验数据,从而使评价数据库的数据更加齐全,给出的信息更加完整。
简介:Legendre矩是以Legendre多项式为核函数的矩,在单位圆内Legendre多项式构成了一个完备正交集。Legendre多项式的正交性使得Legendre矩互相独立,并具有最小的信息冗余度。图像的亮度结构也可以从它的Legendre矩的集合中得以恢复。
简介:AmutuallyorthogonalsystemofrationalSomeapproximationresultsareestablishedfunctionsonthewholelineisintroduced.Asanexampleofapplications,amodifiedLegendrerationalspectralschemeisgivenfortheDiracequation.Itsnumericalsolutionkeepsthesameconservationasthegenuinesolution.Thisfeaturenotonlyleadstoreasonablenumericalsimulationofnonlinearwaves,butalsosimplifiestheanalysis.Theconvergenceoftheproposedschemeisproved.Numericalresultsdemonstratetheefficiencyofthisnewapproachandcoincidewiththeanalysiswell.
简介:Theinitial-boundaryvalueproblemofBurgersequationisconsidered.Aprediction-correctionLegendrespectralschemeisproposed.Itpossessestheaccuracyofsecondorderintimeandhigherorderinspace.Thenumericalexperimentsshowthehighaccuracyofthisapproach.
简介:Pseudospectral(PS)computationalmethodsfornonlinearconstrainedoptimalcontrolhavebeenappliedtomanyindustrial-strengthproblems,notably,therecentzero-propellant-maneuveringoftheinternationalspacestationperformedbyNASA.Inthispaper,weproveatheoremontherateofconvergencefortheoptimalcostcomputedusingaLegendrePSmethod.Inadditiontothehigh-orderconvergencerate,twotheoremsareprovedfortheexistenceandconvergenceoftheapproximatesolutions.RelativetoexistingworkonPSoptimalcontrolaswellassomeotherdirectcomputationalmethods,theproofsdonotusenecessaryconditionsofoptimalcontrol.Furthermore,wedonotmakecoercivitytypeofassumptions.Asaresult,thetheorydoesnotrequirethelocaluniquenessofoptimalsolutions.Inaddition,arestrictiveassumptionontheclusterpointsofdiscretesolutionsmadeinexistingconvergencetheoremsisremoved.
简介:
简介:研究Legendre小波方法求解具有一阶导和二阶导类型的线性Fredholmintegro-differential型方程。应用Legendre小波逼近法把这两类方程分别化为代数方程求解.实例说明。Legendre小波在解决这两类方程时的可行性和有效性.
简介:给出了非线性守恒方程初边值问题的Chebychev-Legendre拟谱粘性法(CLSV).文中,用补偿方法处理边界条件,而对高频部分使用粘性法,以恢复精度.最后证明了在适当条件下,CLSV解收敛于唯一的熵解.
简介:Thefingerjointlinesdefinedasfingercreasesanditsdistributioncanidentifyaperson.Inthispaper,weproposeanewfingercreasepatternrecognitionmethodbasedonLegendremomentsandprincipalcomponentanalysis(PCA).Afterobtainingtheregionofinterest(ROI)foreachfingerimageinthepreprocessingstage,LegendremomentsunderRadontransformareappliedtoconstructamomentfeaturematrixfromtheROI,whichgreatlydecreasesthedimensionalityofROIandcanrepresentprincipalcomponentsofthefingercreasesquitewell.Then,anapproachtofingercreasepatternrecognitionisdesignedbasedonKarhunen-Loeve(K-L)transform.ThemethodappliesPCAtoamomentfeaturematrixratherthantheoriginalimagematrixtoachievethefeaturevector.Theproposedmethodhasbeentestedonadatabaseof824imagesfrom103individualsusingthenearestneighborclassifier.Theaccuracyupto98.584%hasbeenobtainedwhenusing4samplesperclassfortraining.Theexperimentalresultsdemonstratethatourproposedapproachisfeasibleandeffectiveinbiometrics.
简介:Inthispaper,a-posteriorierrorestimatorsareproposedfortheLegendrespectralGalerkinmethodfortwo-pointboundaryvalueproblems.ThekeyideaistopostprocesstheGalerkinapproximation,andtheanalysisshowsthatthepostprocessimprovestheorderofconvergence.Consequently,weobtainasymptoticallyexactaposteriorierrorestimatorsbasedonthepostprocessingresults.Numericalexamplesareincludedtoillustratethetheoreticalanalysis.