Computer recognises archaeological material and fake Van Goghs

June 30th, 2009

People find it very easy to recognise a face, even under very different circumstances. For a computer, on the other hand, it is extremely difficult. Dutch researcher Laurens van der Maaten has developed a new analytical technique which enables the computer to better interpret the content of photos and images, but also of data.

The ‘proof of the pudding’ of his technique for automatic image analysis is a system for the automatic analysis and recognition of archaeological material such as pottery, Roman coins and glass from the Middle Ages. Van der Maaten has also successfully used the technique to distinguish forgeries and paintings by contemporaries of Van Gogh from paintings by Van Gogh himself.

Two major problems alleviated

One of the challenges Van der Maaten faced was the large number of pixels, and thus the high dimensionality of image-space representations. Another major challenge was the variation of images resulting from changes in illumination, rotations or changes of scale of the object. Van der Maaten was able to alleviate these problems by testing new techniques in visualisation experiments and then extrapolating those techniques and re-testing them in a number of variants.

Technique for images and other data

The technique was developed for automatic image analysis in the cultural heritage sector. For example, it can be used for the computer analysis of ancient coins, seeds obtained from archaeological excavations or Van Gogh paintings. Yet Van der Maaten’s research can also be applied to non-visual collections of high-dimensional data, such as the datasets of Statistics Netherlands or the historical radio addresses made by Queen Wilhelmina during WWII.

Collaboration between science and cultural heritage

Van der Maaten conducted his research as part of the CATCH project RICH (Reading Images in the Cultural Heritage), which is geared to automatic image recognition of archaeological objects. The project is a collaboration between Tilburg University and the State Service for Cultural Heritage (Rijksdienst voor Cultureel Erfgoed). The NWO’s CATCH programme includes a further 13 projects in which computer scientists collaborate with a cultural heritage institution on a specific issue.

The research in the RICH project has also led to an advanced arrangement for the digitisation of flint and an application which facilitates the retrieval of archaeological reports.

Source: NWO