中国科学院南京地质古生物研究所机构知识库

Archaeologists often wish to distinguish between groups of cultural artifacts using information collected from descriptions or measurements of their morphological forms. Morphometric methods have played an increasingly large role in such quantitative assessments. However, standard approaches to morphometric analyses are often poorly suited to many artifact types as much variation of interest to archeologists cannot be quantified adequately by sparse sets of landmarks or semilandmarks. Use of measurement conventions also requires that investigators know which aspects of the artifacts under consideration are important to include at the outset of an investigation. In this review results obtained from landmark-semilandmark-based, and image pixel-based assessments of a common set of Paleoindian fluted projectile points are compared. Results confirm that, by itself, PCA is unsuited for the assessment of between-group differences irrespective of data type, but can be useful as a transformation to reduce the dimensionality of a morphological dataset while retaining its effective information content. Landmark-semilandmark data analysed using geometric morphometric methods delivered the lowest-quality results whereas image pixel data analysed by the Naive Bayes machine-learning classifier delivered the highest. Direct analyses of artifact images using geometric morphometric methods delivered very good results. These findings suggest that the direct analyses of digital images and 3D scans, using either geometric morphometric data-analysis methods or machine-learning procedures, can provide archaeologists with tools that improve and extend the scope of their assessments of a wide range of artifact types. (C) 2018 Published by Elsevier Ltd.