Identification of materials based on spectral reflectance is confounded by variations in reflectance magnitude that are independent of the spectral shape. Local variations such as viewing/illumination angles, multiscale soil surface roughness that causes shadows and redistributes light, and soil moisture content, all drive changes in magnitude that are distinct from the spectral variations, and complicate identification and modeling of targets based on spectral features. Normalization metrics that remove magnitude variations can greatly clarify the nature of spectral differences, simplifying interpretation of reflectance features in spectral imagery. Normalized difference measures are particularly useful because of the simplicity of the computation, the convenient scaling, and the ease with which the normalized difference procedure can be extended to multiple dimensions. The two-dimensional normalized difference space described here allows for improved discrimination among bare soils and emergent vegetation when there are multiple soil types in the scene. A 2-D model of soil-specific change in vegetation density is presented. An application of the vector index to mineral identification and mapping is also presented, with an emphasis on band selection.

基于光谱反射率的材料识别被与光谱形状无关的反射率幅度的变化所混淆。局部变化，例如视角/照明角度、导致阴影和重新分布光的多尺度土壤表面粗糙度以及土壤水分含量，都会导致与光谱变化不同的幅度变化，并使基于光谱特征的目标识别和建模变得复杂。消除幅度变化的归一化指标可以极大地阐明光谱差异的性质，简化对光谱图像中反射特征的解释。归一化差异度量特别有用，因为计算简单，缩放方便，以及标准化差分程序可以轻松扩展到多个维度。当场景中有多种土壤类型时，这里描述的二维归一化差异空间允许更好地区分裸土和挺水植被。提出了植被密度中土壤特定变化的二维模型。还介绍了矢量索引在矿物识别和绘图中的应用，重点是波段选择。