Détail de l'auteur
Auteur Roland Eduard Stengele |
Documents disponibles écrits par cet auteur (1)
Ajouter le résultat dans votre panier Affiner la recherche Interroger des sources externes
Kartographische Mustererkennung : rasterorientierte Verfahren zur Erfassung von Geo-informationen / Roland Eduard Stengele (1995)
Titre : Kartographische Mustererkennung : rasterorientierte Verfahren zur Erfassung von Geo-informationen Titre original : [Reconnaissance de modèle cartographique : procédés orienté raster pour la saisie de l'information géographique] Type de document : Thèse/HDR Auteurs : Roland Eduard Stengele, Auteur Editeur : Zurich : Institut für Geodäsie und Photogrammetrie IGP - ETH Année de publication : 1995 Collection : IGP Mitteilungen, ISSN 0252-9335 num. 054 Importance : 147 p. Format : 21 x 30 cm ISBN/ISSN/EAN : 978-3-906513-66-9 Note générale : Bibliographie Langues : Allemand (ger) Descripteur : [Vedettes matières IGN] Photogrammétrie numérique
[Termes IGN] appariement de modèles conceptuels de données
[Termes IGN] base de connaissances
[Termes IGN] base de données localisées
[Termes IGN] données maillées
[Termes IGN] extraction automatique
[Termes IGN] reconnaissance de formes
[Termes IGN] similitude
[Termes IGN] vectorisationIndex. décimale : 33.30 Photogrammétrie numérique Résumé : (Auteur) Geographic Information Systems (GIS) is a key technology in future. Even today this technology is indispensable to a lot of applications for acquisition, management, processing and presentation of spatial data. Usually the acquisition of spatial data is time- and staff consuming and thus an expensive process. Therefore it seems reasonable to open up the information of existing analogue maps and schemes available in great variety for modern GIS-technology. Due to their neutral geometry and thematic contents topographic maps are particularly suitable as basic information for many applications. The scanner technology offers the basis for automating the acquisition of analogue storage media. Management of spatial data beyond the task of pure visualization requires analytic image processing to extract the logical features. This complex problem can be automated by simulating the capabilities of human perception with computers. These activities are focused on the discipline of pattern recognition, a branch of artificial intelligence.
This paper presents an overall view of pattern recognition techniques suitable for cartographic products, subsequently designated as cartographic pattern recognition. A study on the characteristics of this application and the investigations done so far resulted in the requirement to integrate more and more raster-based tools into the recognition-process. Previous approaches are mainly based on a raster-to-vector conversion of images and the patterns are classified by topological and metric features derived from vector-data. However, the analysis of patterns is affected adversely by unavoidable loss of information due to this conversion.
In the field of image processing two raster images are compared by correlation techniques. Template matching is a special correlation technique and describes the search for patterns in the raster image by calculating similarity of the covered image part and a synthetic prototype of a feature class at each position. This technique has to be optimized with regard to two aspects for exposing its potential for cartographic pattern recognition:
1- Higher recognition rates are obtained by calculating similarity, which is robust against small variations of patterns belonging to the same feature class (different size, orientation, thickness of lines etc.).
2- Acceptable calculating periods can only be achieved by a hierarchic strategy.
The measurement of similarity with different weights is of vital importance in regard to both objectives. The basis of this approach is the variable importance of individual pixels for the pattern shape and its identification. High-significant, significant and insignificant pixels of fore-and background are distinguished. The user can assign the weights to the pixels by editing different colors. In this way the acquisition of knowledge describing the characteristics of a template and the small differences between similar templates can be done easily and very efficiently in a graphic manner.
The library of all available templates represents a graphic-designed knowledge base and is far more than a pure database. For instance the method of knowledge-based template matching is suitable for recognition of map text in horizontal orientation. Extensive random samples proved recognition rates of 95.8% for topographic maps of the Swiss Federal Office of Topography on a scale of 1:25'000. Furthermore any cartographic symbols can be located reliable. Above all this is practicable even for symbols, which are not isolated from the graphic context, but also touch or overlap other graphic elements.
Various local raster operators are specifically combined in order to extract the black-colored signature for buildings in topographic maps. 98% of approximately 10'000 buildings can be recognized automatically in a typical map sheet on a scale of 1:25'000.
The development and implementation of a raster-based system for cartographic pattern reco-gnition was a focal research topic. Some special requirements had to be considered. The variety of possible applications for GIS and cartography is indicated by means of several examples: Data acquisition for the establishment of nationwide spatial information systems (e.g. data base of geographic names, project Vector25 of the Swiss Federal Office of Topography), support for cartographic tasks (e.g. map revision, placement of map text) and acquisition of specific data for various GIS-projects (e.g. to document the change of landscape, modeling of objects for generating SD-images of landscape).Numéro de notice : 27136 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse étrangère DOI : 10.3929/ethz-a-001442656 En ligne : http://dx.doi.org/10.3929/ethz-a-001442656 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=56330 Exemplaires(2)
Code-barres Cote Support Localisation Section Disponibilité 27136-01 33.30 Livre Centre de documentation Photogrammétrie - Lasergrammétrie Disponible 27136-02 33.30 Livre Centre de documentation Photogrammétrie - Lasergrammétrie Disponible