How to Better Find Historical Photographs in an Archive - Geographic Driven Reverse Search for Photographs

Historical photographs are important sources for documenting the past of humanity. They can serve as evidence of our society’s history but are also objects of research themselves. Today, the standard format for describing the geographical origin of a photograph is to assign a geoname or point coordinate of the location of the photographer to the image. In view of historical photographs, we are dealing within this operation with spatial uncertainty. In most cases it is not possible to define the exact location of the photographer in form of coordinates in e.g. regions that afflicted strong environmental changes over time. The aspect of spatial uncertainty is also present, when describing the location of the photograph`s origin by a so-called geoname. A geoname can be comprehended to different extents. For example, it can name a region, a city or just a small part of a city. Therefore the question of the exact spatial origin cannot be answered by solely using a geoname. Such aspects of uncertainty in the localization of historical photographs complicate the retrievability in archives for spatially oriented search. Since the photographs cannot be found by entering search queries in the archive, their potential to serve as a data source for research cannot be fully exploited.
From a technical and mathematical perspective, photographs are a two-dimensional representation of information from a three-dimensional space due to its central perspective character. Each object, such as a building or a landscape element, has two-dimensional coordinates in the image and corresponding three-dimensional coordinates in the world coordinate system. This approach describes a reverse geographical search of historical photographs based on this underlying principle. This means, that spatial information will not be defined by one geoname with various possible spatial extents describing where the picture has been taken, but instead we allow to assign global coordinates to each depicted building and object in the respective photograph for a better georeferencing of the whole picture. This method counteracts to the given issue of uncertainty in spatial metadata of historical photographs (Dorner et al., 2018). Map based retrieval techniques will be able to better satisfy the information demand of spatially oriented and object centric scientific disciplines such as archaeology, monument conservation, architecture or landscape planning.
The presentation demonstrates how an object centric search approach could help to better find photographs for the purpose of e.g. spatially oriented humanities. A web-mapping system allows to use web based tools, to georeference objects depicted in a photograph on a dynamic web map. Users can place markers of different types to define objects such as a building, person or landscape element (e.g. mountain, river, forest, etc.) on historical photographs and can then assign them to a real location on a world map. The more items have been located on the map, the more accurate the image origin can be determined and the more information can be assigned to the photograph stored in the archive (visualized in Figure 1).

Figure 1: Visualization of the principle of pasting elements in a photograph and their assignment to world coordinates in a map based on OpenStreetMap data (

https://www.openstreetmap.org

).

An overview map allows to access photograph collections via spatial interests. Photographs shown there can be filtered by entering metadata search queries such as geonames or assigned tags in a search bar (Figure 2).

Figure 2: Screenshot of the overview map with georeferenced photographs, villages and components in the Bavarian-Czech border area (https://photostruk.de)
An additional automated computer-vision-based pre-classification of photographs, helps to filter search results for image categories like e.g. landscape, building, group of persons, portrait. This machine-based image analysis allows a rough categorization of photographs in great archival collections (Eiler et al., 2018), that are not at all or not well documented due to missing manpower or financial resources.
As an outlook, an application in archaeology will be presented. Archaeologists from the University of South Bohemia České Budějovice use the web-mapping tool for historical photographs in order to reconstruct abandoned settlements that were destroyed after the eviction of the German population in the Czech countryside near the national border to Bavaria (Paclíková et al., 2018). The data collected via the online tool will be used for the orientation of the photographs in three-dimensional space. Further on, the oriented images are used for creating 3D reconstructions for the digital preservation of abandoned sites. The online tool has been developed in 2018 in order to serve crowd-sourcing methods to the public in the middle of 2019. The procedure is supported by the European Union within the Cross-Border Cooperation Program Freistaat Bavaria - Czech Republic Objective ETZ 2014-2020.

Bibliography

Dorner, W., Weinfurtner, A. and Graf, S. (2018). Uncertainty in the Spatial Metadata of Historical Photographs: A Geomatic and Photogrammetric Driven Argumentation.
Proceedings of the Sixth International Conference on Technological Ecosystems for Enhancing Multiculturality. (TEEM’18). New York, NY, USA: ACM, pp. 872–877 doi:10.1145/3284179.3284321. http://doi.acm.org/10.1145/3284179.3284321 (accessed 15 November 2018).

Eiler, F., Graf, S. and Dorner, W. (2018). Artificial Intelligence and the Automatic Classification of Historical Photographs.
Proceedings of the Sixth International Conference on Technological Ecosystems for Enhancing Multiculturality. (TEEM’18). New York, NY, USA: ACM, pp. 852–856 doi:10.1145/3284179.3284324. http://doi.acm.org/10.1145/3284179.3284324 (accessed 15 November 2018).

Paclíková, K., Weinfurtner, A., Vohnoutová, M., Dorner, W., Fesl, J., Preusz, M., Dostálek, L. and Horníčková, K. (2018). Geoinformatics and Crowdsourcing in Cultural Heritage: A Tool for Managing Historical Archives.
Agris On-Line Papers in Economics and Informatics,
2: 73–83 doi:10.7160/aol.2018.100207.