Aim. The introduction of the full paper reviews relevant matters and then proposes the NPPNMF feature extraction method mentioned in the title, which we believe is new and effective and which is explained in sections 1 through 3. Our new method incorporates the neighborhood preserving assumption. Section 1 briefs PNMF method. Section 2 explains our new NPPNMF method; it is divided into subsections 2. 1 and 2. 2. Section 3 analyzes the convergence of NPPNMF method; it gives the proof of four theorems. Section4 deals with experimental results and their analysis. Subsection 4. 1 briefs AVIRIS hyperspectral data set. Subsection 4. 3 presents experimental results on such data set in Figs. 2 and 3 and Table 1 and analyze these results. The theoretical analysis in section 3 and the analysis of experimental results in section 4 demonstrate preliminarily that the proposed new method is effective and promising in hyperspectral image classification.%通过对投影非负矩阵分解(PNMF)增加近邻保留假设,提出了一种新的高光谱图像线性特征提取方法——近邻保留投影非负矩阵分解(NPPNMF).NPPNMF保留了高光谱数据在低维特征空间中的局部几何结构,克服了PNMF基于Euclidean的缺点.根据在构造κ近邻图时是否使用训练样本的类标签信息决定了NPPNMF既可以是无监督的特征提取方法,也可以是有监督的特征提取方法,从而提高了PNMF算法的鉴别力.理论证明和高光谱图像数据的分类结果表明了该方法的有效性及应用潜力.
展开▼
