Laser scanner impacts cave archaeology

Figure 1
The prehistoric cave art of Western France is a remarkable record of the very earliest representational artistic expression. The rock art consists of mono and polychrome images of the ice age European fauna, signs and handprints. Stylised, striking and resonant though these images are, their meaning remains elusive despite extensive examination. A British Council supported research project between the University of Bristol and the CNRS in Bordeaux, France is utilising archaeological and computer science expertise to investigate the human perception of the images as they may have appeared at the time of their creation. The chamber context, light conditions, changing qualities of the pigments and the acoustic properties of the cave are all being considered. The goal of this project is to produce a highly accurate virtual computer environment within which the perceptual analysis may be performed. The site chosen for initial investigation is the Upper Palaeolithic site of Cap Blanc in southwest France.

With the latest computer graphics techniques it is now possible to reconstruct archaeological environments and use these to explore different aspects of sites. However, it is essential, if misleading impressions are to be avoided, that any reconstruction should be as accurate and realistic as possible. A fundamental part of any realistic reconstruction is the creation of a highly accurate model of the environment. Where subtle irregularities of the contours of the surface have a significance to the interpretation of an archaeological feature, and where contact with the surface is not recommended, standard surveying techniques may not provide the resolution or accuracy which might be hoped for. To overcome this problem an automated laser scanner (ALS) from Measurement Devices Limited (MDL) was used to acquire the necessary depth date in order to reconstruct a very detailed 3D model of the site.

Cap Blanc

The rock shelter site of Cap Blanc, overlooking the Beaune valley in the Dordogne, contains perhaps the most dramatic and impressive example of Upper Palaeolithic haut relief carving. Discovered in 1909 the site contains a frieze, covering 13m of the wall of the shelter, of horses, bison and deer, some overlain on other images, some carved into the limestone as deeply as 45cms. The sensitivity of the site and the complexity of the carvings precludes acquiring the depth data using traditional intrusive survey and measuring techniques.

One section of the frieze was chosen for scanning in the first instance. The 3x2m sample incorporated one of the most spectacular images of the frieze. This horse, facing right, is almost 2m long and carved in relief between 7cm deep at the mouth and 45cm deep under the stomach. The surface of the carving has eroded away over much of the body and the lower legs and hooves are missing, but on the head the mane, nostril and mouth are clearly visible. A smaller bas relief carving on the jaw of the horse has been interpreted as the head of a bison, and intriguingly above the hind quarters of the horse a small pierced stalagmite has been retained (Figure 1).

Figure 2
An upper and a lower scan of the selected section of the rock shelter were taken with the laser scanner at a grid interval of 2cm. The two scans were downloaded onto a laptop PC on site (Figure 2).

Back in Bristol, the first stage of the visualisation of the survey involved viewing the raw points recorded by the laser scanner.

Figure 3.
To produce a 3-D computer model from the data points, two key issues had to be addressed. First, the data exported from the ALS had to be converted into a triangle mesh, and secondly, the mesh needed to be in a format for displaying graphically.

A program was developed to convert the exported data into a single polygon mesh surface. This program takes as input the depth data from the laser scanner, "stitches" together the different scans into a single "cloud of points", converts them into a mesh of triangles and then displays the resultant wire frame mesh using OpenGL. Finally the triangle mesh is imported into the photo-realistic rendering package Radiance. This requires that a surface be applied to each of the triangles. In this first trial case the same surface was applied to each triangle, the colour was sampled from digital photographs of the cave scene, and the texture was approximated. Quantifiable data describing the actual textures will be recorded during the next phase of this long-term project.

Figure 3 shows the upper and lower scans combined illuminated by a simulated tallow candle in Radiance. The position of the light source is visible in the image, and has a very clear effect of the shadows cast on the scene. A short, 30-frame animation of this scene has been produced, showing a pan across the front of the carving. The motion of this animation helps to demonstrate the 3D aspect of the recording, highlighting the much greater amount of information held in this model than is ever possible in a photograph.

Discussion

The aim of this project was to evaluate the utility of an automated laser scanner to provide detailed data for the development of an accurate computer model of a complex and irregular carved surface. This has certainly been the case and it is now possible to reconstruct complex environments on a computer in a reliable and straightforward manner.

These developments have significant implications for future computer reconstructions of archaeological sites. We are now able to record at high resolution and in 3D the detailed surface structure of a wide range of archaeological features using a simple and relatively inexpensive automated system. Combined with traditional surveys, such scans could provide vital data bases for conservation projects, heritage management, excavation recording, museum presentations, and many other types of archaeological research.