多端元模式下高光谱图像解混的不确定性问题

Analysis of the Uncertainty of Hyperspectral Image Unmixing in Multi-endmember Mode

为了探究多端元解混造成解混结果不确定性的原因,并进一步在解混中克服与降低该不确定性,详细分析了在二分类解混中,解混不确定性的两种表现形式,丰度固定时像元位置的不确定性和像元位置固定时丰度的不确定性,探究解混端元与不确定性的相互作用关系,进而提出一种可降低解混丰度不确定性的端元加权多端元解混方法。实验表明:混合丰度不确定性的存在,同时在保证解混精度的前提下验证了所提出的降低不确定性方法的有效性。

In order to explore the reasons for the uncertainty of the unmixing results caused by multi-endmember unmixing,and further overcome and reduce this uncertainty in unmixing,this article analyzes two forms of unmixing uncertainty in binary unmixing in detail-the uncertainty of pixel position when the abundance is fixed and the uncertainty of pixel position when the abundance is fixed,and explores the interaction relationship between unmixing endmembers and uncertainty.Furthermore,a weighted multi-endmember unmixing method is proposed to reduce the uncertainty of unmixing abundance.The experiment demonstrates the existence of mixed abundance uncertainty,and verifies the effectiveness of the proposed uncertainty reduction method while ensuring the accuracy of unmixing.