A Semi-Automatic Label Digitization Workflow for the Siegfried Map

Laumer, Daniel; Gümgümcü, Hasret; Heitzler, Magnus; Hurni, Lorenz

doi:10.21862/avhm2020.07

Cited by 3 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To properly address the mentioned issues, Laumer et al (2020) assigned each pixel either to a map's foreground (resp. labels) or background with the help of convolutional neural networks.…”

Section: Current Challenges and State Of Researchmentioning

confidence: 99%

Automated Extraction of Labels from Large-Scale Historical Maps

Schlegel

2021

AGILE GIScience Ser.

View full text Add to dashboard Cite

Abstract. Historical maps are frequently neither readable, searchable nor analyzable by machines due to lacking databases or ancillary information about their content. Identifying and annotating map labels is seen as a first step towards an automated legibility of those. This article investigates a universal and transferable methodology for the work with large-scale historical maps and their comparability to others while reducing manual intervention to a minimum. We present an end-to-end approach which increases the number of true positive identified labels by combining available text detection, recognition, and similarity measuring tools with own enhancements. The comparison of recognized historical with current street names produces a satisfactory accordance which can be used to assign their point-like representatives within a final rough georeferencing. The demonstrated workflow facilitates a spatial orientation within large-scale historical maps by enabling the establishment of relating databases. Assigning the identified labels to the geometries of related map features may contribute to machine-readable and analyzable historical maps.

show abstract

“…To properly address the mentioned issues, Laumer et al (2020) assigned each pixel either to a map's foreground (resp. labels) or background with the help of convolutional neural networks.…”

Section: Current Challenges and State Of Researchmentioning

confidence: 99%

Automated Extraction of Labels from Large-Scale Historical Maps

Schlegel

2021

AGILE GIScience Ser.

View full text Add to dashboard Cite

show abstract

“…Supervised machine learning algorithms, such as Deep Convolutional Neural Networks (CNN) have proved to be more effective for such object detection tasks than conventional segmentation methods. CNN-s have been used in various map vectorization applications, such as automatic label extraction (Laumer et al 2020), wetland extraction (Jiao, Heitzler, and Hurni 2020) and improved image segmentation by predictions of areal symbol locations (Groom et al 2020). Saeedimoghaddam and Stepinski used CNN to detect road intersection points and achieved an average of 90% accuracy, with 82% of intersection points extracted (Saeedimoghaddam and Stepinski 2020).…”

Section: Introductionmentioning

confidence: 99%

Automatic vectorization of point symbols on archive maps using deep convolutional neural network

Vassányi

Gede

2021

Proc. Int. Cartogr. Assoc.

View full text Add to dashboard Cite

Abstract. Archive topographical maps are a key source of geographical information from past ages, which can be valuable for several science fields. Since manual digitization is usually slow and takes much human resource, automatic methods are preferred, such as deep learning algorithms. Although automatic vectorization is a common problem, there have been few approaches regarding point symbols. In this paper, a point symbol vectorization method is proposed, which was tested on Third Military Survey map sheets using a Mask Regional Convolutional Neural Network (MRCNN). The MRCNN implementation uses the ResNet101 network improved with the Feature Pyramid Network architecture and is developed in a Google Colab environment. The pretrained network was trained on four point symbol categories simultaneously. Results show 90% accuracy, while 94% of symbols detected for some categories on the complete test sheet.

show abstract

Détection automatique des parcelles sur les plans napoléoniens : comparaison de deux méthodes

Follin¹,

Simonetto²,

Chalais³

2021

revuehn

View full text Add to dashboard Cite

Le cadastre napoléonien fournit la description la plus détaillée du territoire français dans sa globalité au e siècle. Il contient une mine d'informations (structure des parcelles cadastrales, numéro des parcelles, tracé des rivières et des routes, toponymie, bâti…) auxquelles les chercheurs en sciences humaines (archéologues, historiens, urbanistes…) s'intéressent de plus en plus, notamment pour mener des analyses historiques du territoire. Les travaux présentés ici s'inscrivent dans cette perspective et proposent une chaîne de traitement semi-automatique permettant de vectoriser, géoréférencer et assembler des planches scannées du cadastre ancien afin de construire une base de données multiépoques. Dans cet article, nous abordons plus particulièrement les solutions adoptées pour l'étape de vectorisation automatique des parcelles en comparant deux méthodes, l'une basée sur la transformée de Hough probabiliste (THP) et l'autre sur l'algorithme Line Segment Detector (LSD). Après avoir précisé comment ces méthodes ont été implémentées pour notre étude, nous présentons les résultats obtenus, qui montrent la supériorité de la méthode LSD sur la méthode THP pour les planches cadastrales les plus anciennes.The most detailed geographic description of France's whole 19th-century territory is the Napoleonic cadastre. It contains a wealth of information (e.g. cadastre lot structure, lot numbering, rivers and road shapes, toponymy, buildings…) which underpin most historical studies humanités numériques Détection automatique des parcelles sur les plans napoléoniens : comparaison de deux méthodesRésumés of land usage and transformation. We therefore decided to develop a semi-automatic toolchain able to vectorise, georeference and combine scanned old cadastre sheets in order to build a multi-epoch database describing land property. In this article, we focus on the automatic vectorisation of the parcels' shapes and assess the respective merits of two methods we tested: the first based on the Probabilistic Hough Transform (PHT) and the second on the Line Segment Detector (LSD) algorithm. After explaining how we implemented these two methods, we present the results, which show that the LSD approach performs better on the oldest cadastre sheets. humanités numériques spatialisées, imagerie, numérisation spatial digital humanities, digitisation, image processing Entrées d'index MOTS-CLÉS : KEYWORDS: Nous tenons à remercier les anciens étudiants de l'École supérieure des géomètres et topographes, stagiaires du laboratoire Géomatique et foncier, Maïté Fahrasmane, Charlotte Odie et Jean-Marc Beveraggi, sans qui ces travaux n'auraient pu être réalisés. Nous remercions également nos collègues Marie Fournier et Mathieu Bonnefond ainsi que Gilles Berteau de la direction générale des finances publiques de la Sarthe.

show abstract

A Semi-Automatic Label Digitization Workflow for the Siegfried Map

Cited by 3 publications

References 0 publications

Automated Extraction of Labels from Large-Scale Historical Maps

Automated Extraction of Labels from Large-Scale Historical Maps

Automatic vectorization of point symbols on archive maps using deep convolutional neural network

Détection automatique des parcelles sur les plans napoléoniens : comparaison de deux méthodes

Contact Info

Product

Resources

About