Preprocessing of Spectroscopic Data Using Affine Transformations to Improve Pattern-Recognition Analysis: An Application to Prehistoric Lithic Tools
Metadatos
Afficher la notice complèteAuteur
Esquivel Guerrero, José Antonio; Esquivel Sánchez, Francisco Javier; Morgado Rodríguez, Antonio; Romero Béjar, José Luis; García Del Moral Garrido, Luis F.Editorial
MDPI
Materia
Affine transformation Archaeology Flint (chert)
Date
2022-11-13Résumé
The analysis of spectral reflectance data is an important tool for obtaining relevant information
about the mineral composition of objects and has been used for research in chemistry, geology,
biology, archaeology, pharmacy, medicine, anthropology, and other disciplines. In archaeology, the
use of spectroscopic data allows us to characterize and classify artifacts and ecofacts, to analyze
patterns, and to study the exchange of materials, etc., as well as to explain some properties, such as
color or post-depositional processes. The spectroscopic data are of the so-called “big data” type and
must be analyzed using multivariate statistical techniques, usually principal component analysis
and cluster analysis. Although there are different transformations of the raw data, in this paper, we
propose preprocessing by means of an affine transformation. From a mathematical point of view, this
process modifies the values of reflectance for each spectral signature scaling them into a [0, 1] interval
using minimum and maximum values of reflectance, thus highlighting the features of spectral curves.
This method optimizes the characteristics of amplitude and shape, reduces the influence of noise,
and improves results by highlighting relevant features as peaks and valleys that may remain hidden
using the raw data. This methodology has been applied to a case study of prehistoric chert (flint)
artifacts retrieved in archaeological excavations in the Andévalo area located in the Archaeological
Museum of Huelva (Huelva, Andalusia). The use of transformed data considerably improves the
results obtained with raw data, highlighting the peaks, valleys, and the shape of spectral signatures.