简介：在研究了非负矩阵分解（NMF）用于遥感图像融合技术的基础上，改进了一种基于光谱复原的NMF遥感图像融合算法，首先利用矩阵非负性分解求取图像的基向量矩阵，然后综合考虑图像融合后的光谱特性和波形变化情况，对基向量矩阵进行直方图规则化处理，最后得到融合后的图像。实验结果表明，该方法具有较好保持原始图像的光谱信息和空间信息的优点。

简介：ThedeterminationofLuzhou-flavorliquoragesiscarriedoutbythree-dimensionalfluorescencespectroscopycombinedwithnon-negativematrixfactorization(NMF).37samplesofagedliquorswithweightedagesof15,20and25yearswerepreparedbyblendingthreeLuzhou-flavororiginalbaseliquorswithstorageagesof10,20and30yearsindifferentproportions.Thefluorescencespectraofthesamplesweremeasured,andthenfactorizedintobasismatrixandcoefficientsmatrixbymultiplicativeiterativeNMF.Thefluorescencespectra,reconstructedfromthebasismatrix,aresimilartotheoriginalspectra.Thecoefficientsmatrixistakenastheinputofsupportvectormachine(SVM)toestablishapredictionmodelforthedeterminationofliquorages.Comparedwiththeprincipalcomponentanalysis,thepredictionmodelbasedonSVMhasapredictedaccuracybetterthan91.7%.Thismethodcanprovidehelpsforthemarketsupervisionontheagedliquors.

简介：提出了一种基于gamma分布的NMF算法（GNMF）,并将之用于人脸特征抽取.构造了特征子空间,并在特征子空间内实现脸部识别.结果表明,GNMF算法可行且有效,以GNMF为基础的人脸识别率较高.