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

Yekta Said Can,M. Erdem Kabadayi,Petrus Johannes Gerrits

doi:10.1109/access.2021.3074897

Yekta Said Can, M. Erdem Kabadayi + Show 1 more

Open Access

https://doi.org/10.1109/access.2021.3074897

Copy DOI

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With the increased amount of digitized historical documents, information extraction from them gains pace.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.

Highlights

Historical documents are precious cultural sources that help researchers investigate social, historical and economic perspectives of the past
We explored the performance of Unet architecture as the pretrained network for road type detection which has not been used in historical map processing studies
Resnet50 is reported to have better Intersection over Union (IoU) results than Vgg16 and GoogleNet when applied to the historical maps [4]

Summary

Introduction

Historical documents are precious cultural sources that help researchers investigate social, historical and economic perspectives of the past. The digitization of them grants direct access to researchers and the public. Due to maintenance purposes, direct access to these archives could be restricted or not possible. With the help of increased digitization processes in the last decades, they can be analyzed, and researchers can retrieve new information. There are organized and well-funded efforts to make digitized and georeferenced historical maps publicly available including detailed metadata [1]. Automatic processing techniques are applied to the historical maps for information extraction. Intersection of roads, human settlements, forest and flora cover information were extracted in these studies

Objectives

Methods

Results

Conclusion