Font Size:
Low-cost, rapid, mid-resolution 3D data capture using miniature, automatic Unmanned Aerial Vehicles and modern computer vision algorithms
Last modified: 2011-12-20
Abstract
Unmanned aerial vehicles (UAVs), originally developed for and strictly confined to the military and intelligence sectors, have dramatically come down in price and thus increasingly become available for civilian applications. Miniaturisation of guidance and geo-location electronics as well as digital photographic equipment now makes it possible to build fully auto-piloted UAVs weighing no more than 500g, including camera payload, that can quickly produce high resolution digital aerial imagery of small to medium sized areas following pre-programmed flight plans with pre-defined image overlap. At the same time modern computer vision algorithms based on scale-independent feature transforms are capable of extracting 3D point clouds from such matching images and producing geo-referenced ortho-mosaics and terrain models without the need for ground control or terrain models. With ground resolutions of up to 2cm and rapid turnaround times (ca. 1 km² per 30-min. flight, images processed within hours—even straight in the field), miniature UAVs offer an attractive and cost-effective means of capturing mid-resolution 3D and GIS data. Potential applications are in survey and landscape archaeology, particularly when repeated monitoring and/or specific environmental conditions are required for successful data collection (e.g. crop marks) and in areas where availability of government and commercial geo-data is limited. At the site level miniature UAVs can be employed in documenting excavation progress replacing tedious drafting with full-coverage, seamless orthophoto mosaics from which GIS features may be extracted—without any more disruption or delay than a regular photography session. We report on our experiences with these technologies and assess their performance and potential for archaeological applications.
Keywords
UAV; aerial photogrammetry; computer vision; orthophoto mosaics; terrain models