Détail de l'auteur
Auteur R. Brugelmann |
Documents disponibles écrits par cet auteur (1)
Ajouter le résultat dans votre panier Affiner la recherche Interroger des sources externes
Rasterbasierte Methoden zur Gebäudeextraktion aus gescannten Karten / R. Brugelmann (1998)
Titre : Rasterbasierte Methoden zur Gebäudeextraktion aus gescannten Karten Titre original : [Méthodes basées sur les mailles pour l'extraction du bâti à partir de cartes numérisées] Type de document : Thèse/HDR Auteurs : R. Brugelmann, Auteur Editeur : Munich : Bayerische Akademie der Wissenschaften Année de publication : 1998 Collection : DGK - C Sous-collection : Dissertationen num. 504 Importance : 122 p. Format : 21 x 30 cm ISBN/ISSN/EAN : 978-3-7696-9543-4 Note générale : Bibliographie Langues : Allemand (ger) Descripteur : [Vedettes matières IGN] Traitement d'image
[Termes IGN] 1:5.000
[Termes IGN] appariement de formes
[Termes IGN] carte numérisée
[Termes IGN] champ aléatoire de Markov
[Termes IGN] classification
[Termes IGN] détection du bâti
[Termes IGN] données maillées
[Termes IGN] extraction automatique
[Termes IGN] reconnaissance de formes
[Termes IGN] théorie des graphesIndex. décimale : 35.20 Traitement d'image Résumé : (Auteur) Analog maps comprise a large potential of spatial knowledge. Thus they are one of the most important data sources for the creation of digital databases in Geographical Information Systems (GIS). The generation of such digital databases is the bottleneck concerning time and costs of each GIS application. Therefore in this thesis a contribution to a fast and automated extraction of geo-information from maps is presented. The goal of this work is the development of methods for the automatic extraction of areas covered by buildings from the scanned German base map 1:5000 (Deutsche Grundkarte). For this purpose, object and image models of different complexity are formulated. The object classes of the image primitives are found by matching these models.
The analysis of existing approaches in the field of map understanding shows that most of them are based on previously extracted vector data. These vectors only approximately represent the original map data. Metric and topological errors which arise in the course of the vectorization process are frequently complicating a correct interpretation of the data. Therefore raster based approaches will be investigated in terms of their efficiency in map interpretation. Pixel based methods as well as region based ones come into question among raster oriented methods. In this thesis three different raster based approaches for map interpretation are presented.
In a first approach the buildings are detected as hatched areas by investigating parts of run length encoded image rows and columns. In a further step these hatched areas are improved and the borders of the houses are reconstructed by a combination of pixel based mathematical morphology operations and an adapted region growing algorithm. Buildings and Non-buildings are the only object classes.
In contrast, the object model of the second, pixel based approach is more complex and contains more object classes. The interpretation is based on features with their context which are computed for each pixel, and a multivariate statistical classification. A combination of these both approaches in addition to the use of the connected components of the image background is performed and turns out to be successful.
The third, region based approach requires a segmented image which contains node, line and region areas. Markov random fields in combination with the Bayes statistic are used for map interpretation in this approach. These tools enable the classification process to use both the features of the primitives and their spatial relationships. The local neighbourhood relations of the objects are described in the object model by means of cliques of the graph theory. The refined modelling of the map content in this approach allows the extraction of boundaries between adjacent buildings and thus the extraction of single buildings. Results of the pixel based approaches can very well be used as a priori information for the estimation of the object classes. The advantage of the Markov random fields combined with the Bayes statistic is the possibility of formulating a comprehensive stochastic model.
With several tests on real data which are representative for city areas (part of DGK5-sheet 'Karlsruhe Weststadt'), the performance of the developed raster based algorithms for map interpretation is investigated and the results are assessed.Numéro de notice : 46193 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse étrangère Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=58479 Exemplaires(1)
Code-barres Cote Support Localisation Section Disponibilité 46193-01 35.20 Livre Centre de documentation Télédétection Disponible