Automatic Detection of Road Types From the Third Military Mapping Survey of Austria-Hungary Historical Map Series With Deep Convolutional Neural Networks

Abstract

With the increased amount of digitized historical documents, information extraction from them gains pace. Historical maps contain valuable information about historical, geographical and economic aspects of an era. Retrieving information from historical maps is more challenging than processing modern maps due to lower image quality, degradation of documents and the massive amount of non-annotated digital map archives. Convolutional Neural Networks (CNN) solved many image processing challenges with great success, but they require a vast amount of annotated data. For historical maps, this means an unprecedented scale of manual data entry and annotation. In this study, we first manually annotated the Third Military Mapping Survey of Austria-Hungary historical map series conducted between 1884 and 1918 and made them publicly accessible. We recognized different road types and their pixel-wise positions automatically by using a CNN architecture and achieved promising results.INDEX TERMS Convolutional neural networks, digital humanities, digital preservation, document analysis, geospatial analysis, geospatial artificial intelligence, road type detection, image processing.

Automatic Road Extraction from Historical Maps Using Deep Learning Techniques: A Regional Case Study of Turkey in a German World War II Map

“…Can et al [9] identified seven different road types from the Generalkarte historical map series using CNN-based classifiers, with an IoU value of around 0.45 and a pixel-wise accuracy of 0.93. In general, the precision values of different road types were lower than the recall values.…”

“…Manual digitization has still been widely used for reliable labeled data preparation for use in the learning process of artificial intelligence-based approaches. It took Vectorization of historical maps and aerial photographs has been conducted mainly through the on-screen digitization technique, which is time-consuming and labor-intensive [4,8,9]. Manual digitization has still been widely used for reliable labeled data preparation for use in the learning process of artificial intelligence-based approaches.…”

“…Manual digitization has still been widely used for reliable labeled data preparation for use in the learning process of artificial intelligence-based approaches. It took in total 1250 h to digitize and label 300,000 km of roads, including more than 64,000 segments from the Generalkarte, another military map [9]. For the digitization of DHK 200 Turkey, we spent around 1500 h extracting roads for a total of 85 images out of 138.…”

“…In particular, they explained the CNN architectures used in semantic segmentation such as U-Net, SegNet, and DeepLab. Moreover, deep learning-based methods have started to be implemented in historical map-related tasks recently [2,6,9,15].…”

Combining Remote Sensing Derived Data and Historical Maps for Long-Term Back-Casting of Urban Extents

“…Such recent efforts include the mining of (historical) map collections by their content or associated metadata [32][33][34][35][36][37], automated georeferencing [18,[38][39][40] and alignment [41,42], text detection and recognition [43][44][45], or the extraction of thematic map content, often involving (deep) machine learning methods, focusing on specific geographic features such as forest [46], railroads [33,47], road network intersections [48,49] and road types [50], archeological content [51] and mining features [52], cadastral parcels boundaries [53,54], wetlands and other hydrographic features [55,56], linear features in general [57], land cover / land use [58], urban street networks and city blocks [34], building footprints [13,59,60] and historical human settlement patterns [61][62][63]. Other approaches use deep learning based computer vision for generic segmentation of historical maps [64,65], generative machine learning approaches for map style transfer [66,67] or attempt to mimic historical overhead imagery based on historical maps [68].…”