基于Gram行列式的快速端元提取方法

A New Endmember Extraction Method Based on Gram Determinant

高光谱图像端元提取往往涉及到高维空间中单形体体积的计算,使用无需降维的体积公式能够避免信息损失,但却具有极大的计算复杂度。针对这一缺点进行了研究,提出了基于Gram行列式快速的端元提取算法。该算法不需要计算单形体体积,而是利用了体积公式的递推关系,大大降低了计算复杂度。模拟和真实数据试验表明,该算法在保证高精度端元提取的同时,具有极快的端元提取速度。

The endmember extraction methods for hyperspectral imagery generally involve the computation of simplex volume in high-dimension space.The employment of volume without dimensionality reduction though avoids the information loss,suffers from high computational complexity.This paper puts forward a new endmember extraction method based on Gram determinant which is expected to relieve the efficiency issue.The algorithm does not involve the computation of the volumes,instead,it uses the recurrence relationship of volumes which greatly reduces the computational complexity. The experiments on simulated and real datasets verify the accuracy and efficiency of the proposed method.