Heterogeneity characterization is crucial to define the quality of end products and to describe the evolution of processes that involve blending of compounds. The heterogeneity concept describes both the diversity of physicochemical characteristics of sample fragments (constitutional heterogeneity) and the diversity of spatial distribution of the materials/compounds in the sample (distributional heterogeneity, DH). Hyperspectral images (HSIs) are unique analytical measurements that provide physicochemical and spatial information on samples and, hence, are ideal to perform heterogeneity studies. This work proposes a new methodology combining HSI and variographic analysis to obtain a good qualitative and quantitative description of global heterogeneity (GH) and DH for samples and blending processes. An initial step of image unmixing provides a set of pure distribution maps of the blending constituents as a function of time that allows a qualitative visualization of the heterogeneity variation along the blending process. These maps are used as seeding information for a subsequent variographic analysis that furnishes the newly designed quantitative global heterogeneity index (GHI) and distributional uniformity index (DUI), related to GH and DH indices, respectively. GHI and DUI indices can be described at a sample level and per component within the sample. GHI and DUI curves of blending processes are easily interpretable and adaptable for blending monitoring and control and provide invaluable information to understand the sources of the abnormal blending behavior.

中文翻译：

---

基于高光谱图像和变异分析的混合质量评价异质性指标设计

异质性表征对于定义最终产品的质量以及描述涉及化合物混合的过程的演变至关重要。异质性概念既描述了样品碎片的理化特性的多样性（构成异质性），又描述了样品中材料/化合物的空间分布的多样性（分布异质性，DH）。高光谱图像（HSI）是独特的分析测量方法，可提供有关样品的物理化学和空间信息，因此非常适合进行异质性研究。这项工作提出了一种新的方法，将HSI和变异函数分析相结合，以获得对样品和混合过程的全局异质性（GH）和DH的良好定性和定量描述。图像解混合的初始步骤提供了一组混合成分随时间变化的纯分布图，可以对混合过程中的异质性变化进行定性可视化。这些图用作后续变异分析的种子信息，这些变异分析提供了新设计的定量全球异质性指数（GHI）和分布均匀性指数（DUI），分别与GH和DH指数相关。GHI和DUI索引可以在样本级别以及样本中的每个组件进行描述。混合过程的GHI和DUI曲线易于解释，适用于混合监视和控制，并提供宝贵的信息以了解异常混合行为的来源。