ABSTRACT

This study proposes a method for the classification of lithic raw materials by means of hyperspectral imaging, a non-destructive fast analytical technique. The information potential of this approach was tested on a dwelling site dated to mid-late Mesolithic (7200–5800 BP) at Lillsjön, Ångermanland, Sweden. A dataset of lithic tools and flakes (2612 objects) made of quartz and quartzite, was analyzed using a shortwave infrared hyperspectral imaging system. The classification of the raw materials was performed applying multivariate statistical models. A random test set of 55 artefacts was selected, classified according to spectral signature and divided into categories corresponding to different geological materials. The same test set was analyzed with Energy Dispersive X-Ray Fluorescence (ED XRF) to validate the classification. The entire dataset of lithics collected on the site was then classified applying a SIMCA model. The distribution of items on the site was visualized in a 3D GIS platform according to their geological characteristics to highlight patterns that could indicate different use of the space and dynamics of raw materials supply over time.

摘要

这项研究提出了一种通过高光谱成像对岩性原材料进行分类的方法，该技术是一种无损的快速分析技术。该方法的信息潜力已在瑞典Ångermanland的Lillsjön的中石器时代中期（7200-5800 BP）进行了测试。使用短波红外高光谱成像系统分析了由石英和石英岩制成的石器质工具和薄片（2612个对象）的数据集。使用多元统计模型对原材料进行分类。选择了55个文物的随机测试集，根据光谱特征进行分类，并划分为与不同地质材料相对应的类别。使用能量色散X射线荧光（ED XRF）分析同一测试集以验证分类。然后，使用SIMCA模型对在现场收集的岩屑的整个数据集进行分类。在3D GIS平台中，根据项目的地质特征将项目的分布可视化，以突出显示可指示空间使用方式和原材料供应动态随时间变化的模式。