Mapping target signatures via partial unmixing of AVIRIS data: in Summaries
Mapping target signatures via partial unmixing of AVIRIS data: in Summaries
复制标题
DOI：
--
复制
发表时间：
1995-01
期刊：
影响因子：
--
通讯作者：
J. Boardman;F. Kruse;R. Green
中科院分区：
文献类型：
--
作者：
J. Boardman;F. Kruse;R. Green
复杂的AVIRIS场景的完全光谱分解可能并不总是可能的，甚至不是人们所希望的。光谱复杂地区的高质量数据是非常高维的，因此很难完全解开。部分分解提供了一种方法，只解决与调查的特定目标直接相关的那部分数据反演问题。成像光谱学的许多应用可以用以下问题的形式来描述：我的目标签名是否存在于场景中，如果存在，每个像素中存在多少目标材料？这是一个不完全混合的问题。未混合端元的数量比光谱定义的目标材料的数量多一。一个额外的末端成员可以被认为是所有其他场景材质的组合，或者是其他所有材质的组合。一些工作人员已经提出了成像光谱分析数据的部分混合方案，但每个方案在操作应用上都有很大的限制。Farrand和Harsanyi描述的低概率检测方法和Smith等人的前景-背景方法都是这种部分分离策略的例子。这里提出的新方法建立在这些创新的分析概念的基础上，结合了它们不同的积极属性，同时试图绕过它们的局限性。这种新方法在存在任意和未知的光谱混合背景的情况下，对AVIRIS数据进行部分分解，绘制表观目标丰度图。它允许目标材料以丰富的形式存在，从而驱动场景协方差的显著部分。此外，它不需要背景材料光谱特征的先验知识。挑战是找到适当的数据投影，以隐藏背景方差，同时最大化目标之间的方差。
A complete spectral unmixing of a complicated AVIRIS scene may not always be possible or even desired. High quality data of spectrally complex areas are very high dimensional and are consequently difficult to fully unravel. Partial unmixing provides a method of solving only that fraction of the data inversion problem that directly relates to the specific goals of the investigation. Many applications of imaging spectrometry can be cast in the form of the following question: 'Are my target signatures present in the scene, and if so, how much of each target material is present in each pixel?' This is a partial unmixing problem. The number of unmixing endmembers is one greater than the number of spectrally defined target materials. The one additional endmember can be thought of as the composite of all the other scene materials, or 'everything else'. Several workers have proposed partial unmixing schemes for imaging spectrometry data, but each has significant limitations for operational application. The low probability detection methods described by Farrand and Harsanyi and the foreground-background method of Smith et al are both examples of such partial unmixing strategies. The new method presented here builds on these innovative analysis concepts, combining their different positive attributes while attempting to circumvent their limitations. This new method partially unmixes AVIRIS data, mapping apparent target abundances, in the presence of an arbitrary and unknown spectrally mixed background. It permits the target materials to be present in abundances that drive significant portions of the scene covariance. Furthermore it does not require a priori knowledge of the background material spectral signatures. The challenge is to find the proper projection of the data that hides the background variance while simultaneously maximizing the variance amongst the targets.