Semiautomatische Annotation von Orten in digitalisierten Altkarten

Doctoral thesis (2018)

Old maps are increasingly being digitised and included in digital libraries to provide access to these historically valuable documents to a wider audience. The need was identified to enrich the digital ... [more ▼]

Old maps are increasingly being digitised and included in digital libraries to provide access to these historically valuable documents to a wider audience. The need was identified to enrich the digital copies of the old maps with annotations that enable a quick search for relevant information and linking with other documents. Because such a manual annotation of old maps is a labour- and cost-intensive process and is therefore only carried out for particularly valuable documents, the large mass of old maps remains unexplored. With this motivation, the aim of this research project was to automatise to a great extent the process of object detection and annotation in the old maps. However, existing methods for automated information extraction from maps are designed for modern maps. They are usually limited to certain map types that have similar graphical characteristics. Alternatively there are tools for manual annotation of map data. Until now, annotation of old maps was only supported by tools that enabled manual extraction of map data and the digital storage of annotated information. The research objectives for this dissertation have been formulated based on the research frame given at the beginning of this work. The research objectives include automatic identification of place markers, automatic detection of text labels, their Optical Character Recognition (OCR) and an automatic linking with a modern map or a geographical database. In order to meet the research objectives, mainly methods of machine learning were used. Both classic machine learning methods and deep-learning-based methods have been extended to meet the requirements of the specific domain of old maps. Old maps of Luxembourg and the Franconian Circle were used as a data set. In order to find place markers automatically, a template matching variant was developed which is robust against interfering objects next to the place markers to be recognised. The method is also applicable in other areas with similar properties. Furthermore, both rule-based and statistical methods were tailored to the domain of the old maps with the purpose to find the position of the text labels. In order to make the identified texts readable, a new OCR engine was implemented that produces only a quarter of the character errors compared with a state-of-the-art OCR tool, while using the same training data. The georeferencing was examined for two types of map pairs: an old and a modern map, and two old maps. It was found that automatic georeferencing between two old maps can not only be used to find errors in manual annotations, but can also provide new historical insights, such as small changes in the copper plates from which the maps were printed (e.g. adding a small number of places that is not documented anywhere else). A referencing and annotation tool has been implemented to collect training and test data. This tool can also be used to test the automations and to correct automatically collected data. Overall, the present work makes a contribution in the areas of object recognition in graphics, scene text detection, OCR and georeferencing. It is shown which effort is needed to make individual parts or the entire area of old maps automatically readable and where the current limits of the automation are. Finally, this research shows the huge potential and the impact of such digital tools in the entire area of digital humanities. [less ▲]

Georeferencing of Place Markers in Digitized Early Maps by Using Similar Maps as Data Source

in Digital Humanities 2017: Conference Abstracts (2017)

Annotating and Georeferencing of Digitized Early Maps

Poster (2016, July)

Original early maps are usually only accessible for a small group of researchers and librarians because they are very old and sensitive, and could be easily destroyed. However, they are a valuable ... [more ▼]

Original early maps are usually only accessible for a small group of researchers and librarians because they are very old and sensitive, and could be easily destroyed. However, they are a valuable knowledge source for historical research, because they are also political and cultural evidences of its time. In the age of Digital Humanities, online access and information search in digitized historical documents and early maps allows people from all over the world to work with such artefacts of cultural heritage. However, the digitization solely generates images of the artefacts without any access to the semantics of the documents. [less ▲]

RAT: A Referencing and Annotation Tool for Digitized Early Maps

Scientific Conference (2016, June)

RAT is designed to support users in identifying place markers in digitised early maps and to link these place markers to modern maps. RAT facilitates a geo-referencing by suggesting the most likely modern ... [more ▼]

RAT is designed to support users in identifying place markers in digitised early maps and to link these place markers to modern maps. RAT facilitates a geo-referencing by suggesting the most likely modern places based on an estimated mapping and a phonetic search for place names. [less ▲]