Abstract: Due to the lack of priori knowledge, it is difficult to determine the number of endmembers for hyperspectral imageries. Aiming at this problem, this paper proposes a new method of estimating virtual dimensionality, its results can be used to estimate the number of the endmembers. Based on the principle of Orthogonal Subspace Projection(OSP), this method extracts and separates the endmember signals, estimates the number of endmembers by comparing the residual value and threshold. Experimental results prove that it is feasible to estimate the number of hyperspectral and PHI hyperspectral image data and it is more flexible and accurate than Nerman-Pearson method.

Key words: hyperspectral image, Orthogonal Subspace Projection(OSP), virtual dimension

摘要： 针对在缺乏先验知识的情况下难以确定高光谱影像端元个数的问题，提出一种新的虚拟维数估计方法，其结果可作为端元个数的估计。该方法采用正交子空间投影(OSP)原理，逐个提取并剥离端元信号，通过比较残余值与阈值，实现虚拟维数的估计。对模拟高光谱数据和PHI高光谱影像数据的实验结果验证方法的可行性，与Nerman-Pearson法相比其具有更高的灵活性和准确性。

关键词: 高光谱影像, 正交子空间投影, 虚拟维数

CLC Number: 

• TP393