Forschung
Research questions
Questions about urban development and its integration into the surrounding area and beyond are at the forefront of the investigations.
Approaches and Methods
UAVs ("drones") in combination with photogrammetry methods have established themselves as a standard tool in archaeological documentation. Multicopters with good cameras are becoming safer and cheaper nowadays. In contrast to stereo-based approaches, single camera systems are often based on the principle of "structure from motion". In terms of software, in addition to many commercial products, there are also freely available packages such as VisualSFM and open source products such as MicMac, OpenMVG that can deliver comparable results. However, dealing with the free products is often a major challenge. This is especially true for the customisation of the free parameters, which offers the user maximum freedom on the one hand, but can lead to unsatisfactory results on the other. The Archaeocopter project has been dealing with the handling and analysis of existing free software packages in the context of archaeological documentation since 2013. New software modules and alternative solutions are also freely available and are constantly being developed. As part of this project, a semi-automatic software called Archaeo3D was developed to optimise and control the entire reconstruction process. The resulting 3D models are heavily dependent on the quality and correlation of the input images. Videos and photos are automatically imported and processed. The software is able to rearrange or adjust the processing modules and adjust the parameters depending on the current hardware and the shooting situation and complexity. VisualSFM, COLMAP and CMPMVS are the main tools used by Archaeo3D. To create a 3D model, standardised steps are first used to create a set of planned, high-resolution photographs, which are then photogrammetrically stitched together. This is in contrast to videogrammetry, which has its origins in robotics for localisation and recognition in real-time systems. However, with the Jkeyframer we have a tool that allows for two solutions:
- A minimum image set for a rough 3D model is provided, showing that the data is sufficient to compute a full 3D model.
- An optimal image set is provided to create a high quality 3D model.
In addition, an optimal solution to the "min-max problem" is sought. It has been found that the videogrammetry approach is more robust and faster, more flexible and not as error-prone as the classical photogrammetry approach. Since modern cameras store the positions of captured photos, usually with so-called geotags in EXIF format, georeferencing can be done automatically. This is not yet readily possible with videos, requiring additional manual georeferencing.
While the methods presented have their strengths and weaknesses, an initial assessment of simple hill visualisations already proved the high potential of the collected data. The digital elevation models (DEMs) derived from the UAV images by the web service DroneDeploy allow to distinguish even the faintest remnants of human settlement activities such as eroded walls, platforms, ditches and fields. However, the DEMs processed with DroneDeploy also have some quality issues. Many artefacts are visible on the DEMs, which sometimes make interpretation difficult. Experimental tests with other structure-from-motion algorithms have shown that better results are possible with data collected in the field. This would improve the basis for applying the visualisation and image processing methods described above.
In addition, an optimal solution to the "min-max problem" is sought. It has been found that the videogrammetry approach is more robust and faster, more flexible and not as error-prone as the classical photogrammetry approach. Since modern cameras store the positions of captured photos, usually with so-called geotags in EXIF format, georeferencing can be done automatically. This is not yet readily possible with videos, requiring additional manual georeferencing.
While the methods presented have their strengths and weaknesses, an initial assessment of simple hill visualisations already proved the high potential of the collected data. The digital elevation models (DEMs) derived from the UAV images by the web service DroneDeploy allow to distinguish even the faintest remnants of human settlement activities such as eroded walls, platforms, ditches and fields. However, the DEMs processed with DroneDeploy also have some quality issues. Many artefacts are visible on the DEMs, which sometimes make interpretation difficult. Experimental tests with other structure-from-motion algorithms have shown that better results are possible with data collected in the field. This would improve the basis for applying the visualisation and image processing methods described above.
Research history
The nomadic empires of Central Asia are particularly interesting phenomena in the history of civilisation. Despite the small population and economy of the Eurasian steppes, they had an enormous impact on the course of human history. Although these empires relied on the military strength and mobility of a predominantly nomadic population, they repeatedly developed significant urban centres. For about twenty years, the German Archaeological Institute, the Mongolian Academy of Sciences and the National University of Ulaanbaatar have been jointly researching urban settlements in a nomadic environment, their characteristics and intercultural influences in Mongolia's Orkhon Valley.
Research objectives
Our aim is to create a comprehensive corpus of high-resolution cultural landscape data in the Karabalgasun and Karakorum area and the entire Orkhon Valley to provide a more holistic view of settlements and land use in the urban sites of the Orkhon Valley. Efforts are being made to improve existing visualisations and to create meaningful images of anthropogenically influenced landscapes that preserve the positive characteristics of the individual technologies applied. The project also focuses on the further development of methods.