Spectroscopic Imaging for Cultural Heritage: Data processing strategies for XRF and Hyperspectral imaging

Datum

10.02.2017

Zeit

10:30 - 11:30

Sprecher

Dr. Matthias Alfeld

Zugehörigkeit

Sorbonne Universités Paris, Laboratoire d’archéologie moléculaire et structurale, LAMS

Sprache

en

Hauptthema

Physik

Andere Themen

Physik, Gesellschaft, Philos., Erzieh., Umwelt

Host

Silke Merchel, Jürgen Fassbender

Beschreibung

The technological progress in the last decades has allowed to expand spectroscopic investigations from spot analysis to imaging and to become a common method in recent years. Few fields of science benefit from this more than the investigation of cultural heritage objects. This is due to their unique nature, which requires non-destructive investigations, and their high heterogeneity on the macroscopic and microscopic length scale, which complicates the acquisition of representative results. The two most prominent spectroscopic imaging techniques are X-Ray Fluorescence (XRF) imaging, which yields elemental distribution images, and hyperspectral imaging in the visible and near infrared range with several hundred color channels, yielding images with chemical contrast. The investigation of cultural heritage objects is often only possibly in-situ, i.e. at the excavation site or in a museum/gallery. Here the conditions are less controlled and less well-defined than in laboratories. At the same time the experimental time is limited, as the stays for in-situ experiments cannot be indefinitely extended. So, data processing faces two different challenges: At the one hand, data needs to be rapidly inspected in-situ to control its quality and to guide further measurements. At the other hand, after the experiment the data needs to be carefully evaluated in order to not miss details in the data. Furthermore, artifacts in the data due to the challenging experimental conditions need to be taken into account. In this talk I will present how matrix factorization techniques with non-negativity constrains support the interpretation of spectroscopic imaging for both challenges. Further, we will show how 3D models of the investigated object acquired by photogrammetry allow to correct for artifacts resulting from the complex shape of an object during XRF imaging and discuss the benefit of Artificial Neural Networks in imaging spectroscopy. The validity and value of these approaches will be demonstrated on spectroscopic imaging data acquired on polychrome pieces of the Frieze of the Siphnian Treasury at the Delphi sanctuary in Greece (approx. 525 BC) and the polychrome decoration of Egyptian Tombs from the time of Ramses II (c. 1303 BC – 1213 BC).

Letztmalig verändert: 31.07.2017, 17:25:15