MOBIMA – Mobile Imaging in Archaeology http://www.idesam.umu.se/english/research/research-projects/mobima--mobile-imaging-in-archaeology/

Project description

Background and problems
Archaeology needs chemical and spectroscopic analysis for a detailed characterization of materials and objects. This characterization should be done with fast non -destructive screening methods. One can thus examine large areas (many samples and observations) during limited time without affecting the examined object / material. Chemical analysis based on near infrared (NIR) is a growing field with a clear potential for the future (Grahn & Geladi 2007). By generating hyperspectral images, we will be able to cover large areas very quickly and each image pixel contains via spectroscopy chemical information and we will then be able to determine contrasts, gradients, classes, and also concentrations. At the same time, we maintain the strength of a traditional documentation in the form of images.
We want to test and investigate rapid non -destructive visualization based techniques in the laboratory and in the field suitable for archaeological research. By taking pictures of many wavelengths (a hundred or several hundred) in chemical and biological informational wavebands (near infrared), one can quickly get a snapshot of a situation that would otherwise require hundreds of time-consuming and expensive laboratory analysis. This is a new and innovative technology with potential interest to archaeologists, art historians, chemists, computer scientists and statisticians and will provide new knowledge in archaeological research.
Within analytical sciences and technology (eg, analytical chemistry) there is a trend of taking the analytical method to the sample instead of taking the sample to the instrument. This is a well-known procedure in process analytical technology and is used in pharmaceutical-, pulp- and paper industry. Our idea is to study situations that require hyperspectral image analysis in Archaeology by taking high-tech analytical images with spectral information in the near infrared wavelength range of different situations in archaeological contexts in order to obtain chemical information. Furthermore, we want through advanced data analysis (1) identify areas of interest (2) extract these areas (3) further analyse these phenomena in detail in order to allow for in-depth interpretations of visual information. The methodology will be tested in practice and will be tested and evaluated in different field situations. Conventional NIR spectroscopy and wet chemical analyses will be used and integrated in an optimized way.
A typical example of the area of interest in Archaeology is human use of colour and pigments, in particular the so-called red ochre. Human manipulation of iron oxide occurred in northern Scandinavia already during the Stone Age and today there are visible remains of at paintings on rock surfaces or as a red colourings in soils in prehistoric dwellings. It is not fully known how the ochre was produced, although several hypotheses exist. Evidently, there is a need for technical analysis (and in particular those that provide chemical information) of these materials to provide new knowledge about the phenomenon.
Another theme are the chemical traces that prehistoric humans left behind, that have retained in soils and sediments, and that can be analysed after a considerable amount of time. Here, image analysis and NIR new capabilities to produce interpretable information related to spatial organisation, human impact, etc.
These studies can be used in archaeological sites studies as well as in post archaeological work on excavated materials. There are also opportunities for further studies where there is a need for non -invasive analysis of archaeological materials. Also, museums hold large amounts of materials with good analytical potential. Previously, there has been limitations in analytical possibilities of these materials because of the destructiveness of standard techniques but NIR-technology will give new opportunities in this respect.

Project Goals
MOBIMA aims to replace certain time consuming expensive laboratory analyses by applying quick visual image analytical information in archaeological research and education.
MOBIMA will test and apply hyperspectral imaging systems (VIS and NIR) as rapid visual techniques to be used in archaeological studies. The idea is that the technology should be able to use in forensics, crime scene investigations and environmental studies.
Within MOBIMA we also want to develop a course and contribute to interdisciplinary education in the IT field. We seek here for a close relationship between research and education, both at postgraduate level at an advanced level.
MOBIMA want to contribute a new methodological developments in Archaeology. NIR technology has so far been used very little for archaeological applications.
Data generated within MOBIMA will result in a database of spectral information over several different archaeological phenomena to enable further research.

Questions to answer:
- Which sample types are suitable to investigate the field and/or in the laboratory?
- Which classifications can be done and with what accuracy?
- How is the relationship between image data and external reference analysis?
- How to transfer results achieved laboratory to field situations?
- How to statistically evaluate and validate the reliability of methods?
- How can you improve exchange of prehistoric rock images, thereby improving the possibilities of interpretation, ie by producing information that is not visual?
- How can we improve information exchange and understanding of soil and sediment profiles in the context of archaeological sites by using NIR hyperspectral images?
- How can this be integrated into education/training?
- How is knowledge spread to the public through IT technology?

Methodology
Archaeological and environmental archaeological investigations
An archaeological survey site is usually a complex environment to survey, collect and organize information from. We will apply image analysis techniques with hyperspectral visualization / analysis of different types of materials from archaeological investigations, primarily of the type where the Near-infrared spectral information available, sediments of different types, bone, etc.
Here we want to work with the ability to both record and analyse any field situation an archaeologist may face. This technique can also be seen as a prospecting tool (i.e. to find out the conditions for) where as little damage as possible is initially done to prehistoric remains on interest, but one still maintain the ability to evaluate information and ask new questions or choose new survey strategy before further, often destructive, investigation begins.
Hyperspectral imaging
A hyperspectral image is a digital image which include 256x320 pixels in which each pixel can represent a spectrum with a large number of wavelengths, let’s say 256. This provides a great data cube of 256x320x256 = 20000000 number that was composed of 2 or 4 bytes. Acquiring such pictures is made possible by the newly developed detectors that are built of InGaAs (900-1700 nm) or HgCdTe (1000-2500 nm). A spectroscopic filter mechanism allows selection of wavelengths. This is done quickly so a total hyperspectral image can be made in seconds or worst few minutes. The technique is non -destructive. Also, the near-infrared spectral enables among others: Classification of organic and biological materials and also to some extent inorganic materials. Calibration of a material can lead to concentration images and interpretation of the spectral information. The advantage of this to analysing a small sample of an object which then partially destroyed, and then having to wait several weeks for results of a wet chemical analysis.

Within this project, we have access to a number of analytical tools that can create hyperspectral images in the laboratory. We will test these with a large number of archaeological samples that have not been investigated in this way before. Furthermore, we will test the ability to use NIR-instruments in the field for rapid on-line measurements as diagnostic instrument during archaeological surveys/ excavations. This enables faster and more reliable data for the entire archaeological approach. Typically, these decisions after completion of the field survey / received assay results. Since Archaeology is a destructive method in relation to its source material, it is important to better decisions is early in the investigation process.

Multivariate Data Analysis
The amount of information in a hyperspectral image is overwhelming and therefore requires effective methods for data reduction, analysis and visualization. Spectral data in each pixel (pixel) can give rise to interactive data exploration, classification and calibration against external data. The software enables, while analysing the data, also a statistical interpretation. We will use a software that can be run on a personal computer by normal high standard and providing interactive visual interpretations. One can e.g. create an image with classes in different colours showing unaffected / painted areas of a rock painting where it can be very difficult to visually assess motives. Furthermore, in studies of soil and sediment visualize a gradient from a natural background to the more heavily cultural influenced areas such as dwelling site studies in Archaeology.
Commercially available and already proven products will be used (Evince; UmBio 2012). Chemometric analysis to the identification, classification and quantification of the archaeological and environmental archaeological materials that are considered to be of interest. Comparisons between the reference analysis and image analysis are of central importance.

Reference Analyses
Obtained NIR spectra analysed from sediment / soil wet chemical will be related to phosphate analysis, magnetic susceptibility and trace element determinations (ICP-MS/XRF). During field work, field-based XRF devices (XRay fluorescence) will be used, primarily with respect to the major components. In Archaeology, analysis of soil phosphate content and magnetic susceptibility are well established exploration methods (Linderholm 2010) that gives a good insight into the impact of humans on their environment that are thus well suited to be related to the NIR spectra.