Application of Sky-View Factor for the Visualisation of Historic Landscape Features in Lidar-Derived Relief Models

Article excerpt

Introduction

Airborne laser scanning (also known as lidar imaging) has provided major advances in remote sensing. Because the technology enables production of digital elevation models of very high resolution (better than l m) with high relative elevation accuracy (centimetre level), even under forest, it has become well established in archaeological applications (Bewley et al. 2005; Devereux et al. 2005).

However, effective interpretation of digital elevation models (DEM) requires appropriate data visualisation. Analytical relief shading is used in most cases, but although several authors describe the advantages of sophisticated visualisation (Devereux et al. 2008; Doneus et al. 2008), it remains limited to multiple angle shading and elevation data subtraction. Devereux et al. (2008: 477) stress that

'... any image product which removes the directional problems of hill-shading would be of great value for archaeological survey.'

This paper addresses the problem by introducing sky-view factor (SVF), a visualisation technique based on diffuse light that overcomes the directional problems of hill-shading.

Applying the SVF for visualisation purposes gives advantages over other techniques because it reveals small (or large, depending on the scale of the observed phenomenon and consequential algorithm settings) relief features while preserving the perception of general topography. Rather than just presenting or visualising the same information in a new way, it extracts new information that can be further processed. In addition to the studies of the past cultural and natural landscapes it can be effectively used in other scientific fields in which digital elevation model visualisations (Yoeli 1965; Brassel 1974; Horn 1981; Imhof 1982; Kennelly 2008) and automatic feature extraction techniques (Kweon & Kanade 1994; Lopez et al. 1999; Wladis 1999; Kim et al. 2005; Briese et al. 2009) are indispensable, e.g. geography, geomorphology, cartography, hydrology, glaciology, forestry and disaster management.

Terrain visualisation in archaeological investigation

Standard analytical hill-shade is widely used for archaeological interpretation because it is a plastic and illustrative representation of the topography (Figure 1A). It is easy to compute and easy to interpret. The surface is illuminated by a direct light. An imaginary light source ar an infinite distance, with a constant azimuth and zenith angle for the entire studied area, is used to compute the incidence angle of the light on the relief surface. Areas perpendicular to the light beam are illuminated the most, while areas with an incidence angle equal or greater than 90[degrees] are in a shade. A greyscale colour table is usually used for its visualisation, because the colour change from white through grey to black enhances the perception of the relief morphology. However, this limits the visualisation--especially in dark shades and brightly lit areas, where no or very little detail can be perceived. A single direction of the light beam also fails to reveal linear structures lying parallel to it (Figure 2).

Illuminating a surface from multiple directions, e.g. equally spaced between 0[degrees] and 360[degrees], enhances the contrast but is inefficient at best for larger surveys and may lead to inconsistencies in the interpretation. Hill-shaded images are frequently used to guide ground surveys because comparing multiple images in the field is extremely inconvenient. Combining multiple shading layers by considering only the mean or the maximum of an individual pixel represents a step towards an improved understanding of the results (Figure 1B). Even though this method is less appropriate for automatic edge extraction than e.g. the Sobel filter (Figure 1E), it is favourable in archaeological interpretations for it reveals, not only the most perceptible edges, but also the more subtle ones.

Because images created by illumination from several angles are highly correlated (the same scene is viewed), it is possible to 'summarise' information by a mathematical transformation with principal component analysis (PCA) (Devereux et al. …