Taking vegetation marks into the next dimension(s): Mapping the hilltop settlement of Montarice by a multi-dimensional analysis of aerial imagery

Abstract

This paper examines the multi-dimensional potential of archived analogue and digital aerial archaeological photographs. First, such an image dataset holds several inherent spectral dimensions which can be explored. Second, there are also three geometrical dimensions stored in the images. Comparing sequences of images over time even delivers an additional time component. More in particular, aerial imagery acquired over the sites of Montaricehill and ColleBurchio (Regione Marche, Italy) during the Potenza Valley Survey (PVS) will form the subject of this talk. Both promontories are to be situated immediately north of the present course of the Potenza river. The sites have a long settlement history, starting at least in the Middle Bronze Age for Montarice, and probably in the Roman Late Republican period for ColleBurchio. Next to the overwhelming amount of artefacts that have been collected during PVS artefacts surveys over the years, and the successful full coverage geophysical prospections organised here, both sites have also repeatedly revealed themselves in terms of interesting vegetation and soil marks. The main idea of this paper is to examine all the imagery taken since the start of the PVS and explore the potential of state-of-the-art image-based modelling (IBM) techniques in reconstructing the three-dimensional geometry of the photographed hilltop sites. Since dense image matching – as a part of IBM – allows to model the canopy of the vegetation at the moment of the photographic survey, it can yield a three-dimensional geometrical representation of the plants. The latter can be the focus of its own information extraction process, using techniques mainly developed in the fields of computer graphics and airborne laser scanning. However, the true power lies in the combination of the spectral and geometrical dimensions as this approach can even overcome hurdles that are typical for each information set on its own. In other words: the newly extracted geometry will be used to enhance the spectral dimensions of the imagery, hereby increasing the visibility and readability of the imaged vegetation (and soil) marks.