34229

Open reference system. Hyperspectral and NIR data in a geological library for archaeology.

Johan Linderholm 1 ; Paul Geladi 2 ; Claudia Sciuto 3
Favorite this paper

Introduction
The lecture is about the exploration of a database containing NIR and NIR imaging data of inorganic material of important bedrock and mineral types. Characterization of archaeological lithic materials is important to understand the dynamics of geologic resources exploitation in ancient societies. Provenance studies of artefacts shed a light on past trade networks and economical systems. The characterization is traditionally mostly done on samples in laboratory applying destructive techniques. The MOBIMA (mobile imaging in archaeology) project’s purpose is to experiment non-destructive fast methods for lithic characterization.
Material
Within the project several data items have been investigated, both geological reference collections and in-field studies of bedrocks in 9 sites in north Scandinavia. The dataset contains Hyperspectral images and Near Infrared spectra of about 70 stones and 50 minerals from the geological reference collection at Umea University. The combination of Hyperspectral imaging and NIR spectra for each sample contributes to the creation of a robust data background. The data collected during future researches will be matched to the existing dataset to spot correspondences and similarities. This computation, using chemometric methods, allows the identification of artefacts measured directly in the field (with a portable NIR probe) or in the lab. The library is a first step for the elaboration of an on line matching system for classification and research on lithic materials.
Experimental
To collect the data a MicroNIR 1700 infrared system (JDSU) with a spectral range of 908.1-1676.0 nm (128 bands). The probe uses illumination via two integrated vacuum tungsten lamps and a spectralon white reference was used. The Hyperspectral images have been produced with a sisuCHEMA pushbroom shortwave infrared hyperspectral imaging system (Spectral Imaging Ltd, Finland) for acquiring images (320 × 430) from 1000 to 2498 nm at intervals of 6–7 nm (255 bands), also using spectralon white reference. The images obtained with this instrument were transformed into pseudo-absorbance using Evince image analysis software. All data are pre-processed by the standard normal variate (SNV) in order to correct for light scattering, and centered prior to principal component analysis (PCA).
Results
The chemometric analysis of MicroNIR spectra shows no outliers and the interpretation of major bedrock and mineral groups is possible. The Hyperspectral images were treated in the same way using the image average with similar results but with the benefit of being able to handle individually heterogeneous specimens. The reference system statistics is built on the comparison between the two PCA (NIR and HI) and bedrock specimen geological classification.
Conclusion
Several spectral libraries are already available free from various agencies. The USGS has a database available at http://speclab.cr.usgs.gov/spectral-lib.html. This library contains minerals, man-made materials, vegetation and other materials. Jet Propulsion Laboratory has ASTER Spectral library version 2.0 available at http://speclib.jpl.nasa.gov. Also, the Swedish SGU provides SWIR images on bedrock drill cores (http://www.sgu.se/en/). The MOBIMA dataset is going to be released on line and Open Access as a part of the SEAD (Strategic Environmental Archaeological Database http://www.sead.se/). For available files metadata about product’s detail are included.