Computer vision-based orthophotomapping of complex archaeological sites: The ancient quarry of Pitaranha (Portugal-Spain)

Abstract

Ancient quarries are intriguing archaeological sites, but their detailed recording is difficult. The complex morphological and topographical characteristics of the quarry landscape at Pitaranha (border Portugal–Spain), as well as the severe modification of the terrain configuration by both intensive quarrying and the intricate logistical extraction infrastructure make a survey of this site rather complicated. Since an accurate digital representation of the topographical surface is elementary for the spatial analysis of the quarry and the availability of an orthophoto map a necessary prerequisite for fast and effective site navigation, the acquisition of such information is a crucial component of efficient quarry research. In situ field work also benefits from a detailed and accurate orthophoto map, while a digital three-dimensional (3D) model constitutes a fundamental layer in GIS-based spatial analysis studies. For complex sites such as the quarry of Pitaranha, this is a difficult, timeconsuming and very expensive operation. In this poster, a cost-effective approach is presented for the mapping of the Roman quarry site of Pitaranha (Portugal–Spain) based on recent technological advances in computer vision and photogrammetry algorithms. Using a radio-controlled digital reflex camera attached to an unmanned low-altitude aerial photography system (Helikite-based), aerial photographs were acquired that allowed for the acquisition of the necessary low-altitude aerial footage in very unstable wind conditions above the quarry. Using a semi-automatic Structure from Motion (SfM) approach complemented with Multi-View Stereo (MVS) algorithms, the aerial images were processed for the production of an orthophoto map and Digital Surface Model (DSM) of the site. An enormous advantage of the presented approach is the minimal prerequisites: for example, no calibrated cameras and optics are necessary during the image acquisition stage. Besides focusing on the acquisition and processing method, the poster evaluates the accuracy of the generated products. For the Pitaranha quarry, the horizontal and vertical accuracy at the 95 % confidence level were 13.7 cm and 31 cm respectively. Given the unordered image collection of vertical and oblique aerial photographs that are characterised by a maximum ground sampling distance of 8 cm and acquired with a lens that suffered from quite some distortion, the reported positional accuracy of these data sets is considered very good (planimetric) to good (altimetric), and satisfactorily accurate for 1:200 hard-copy mapping.