Centre de documentation
scientifique

Ein Verfahren zur Rekonstruktion von dreidimensionalen Objektmodellen aus digitalen Bilddaten / B. Straub (1991)

Public Titre : Ein Verfahren zur Rekonstruktion von dreidimensionalen Objektmodellen aus digitalen Bilddaten Titre original : [Un procédé pour reconstruire des modèles d'objets en 3D à partir d'images numériques] Type de document : Thèse/HDR Auteurs : B. Straub, Auteur Editeur : Munich : Bayerische Akademie der Wissenschaften Année de publication : 1991 Collection : DGK - C Sous-collection : Dissertationen num. 380 Importance : 122 p. Format : 21 x 30 cm ISBN/ISSN/EAN : 978-3-7696-9426-0 Note générale : Bibliographie Langues : Allemand (ger) Descripteur : [Vedettes matières IGN] Photogrammétrie numérique [Termes IGN] base de connaissances [Termes IGN] compensation par faisceaux [Termes IGN] extraction de traits caractéristiques [Termes IGN] niveau de gris (image) [Termes IGN] primitive géométrique [Termes IGN] reconstruction 3D Index. décimale : 33.30 Photogrammétrie numérique Résumé : (Auteur) In this dissertation, an approach for the reconstruction of 3-dimensional object models analyzing digitized images of > 2 different perspectives is presented. The reconstruction process works without any explicitly defined a priori information about the object and the acquired redundant image information is utilized to recognize and eliminate reconstruction errors. The strategy is derived from a so-called Tetrapoden-model as a possible interpretation of human visual perception. The System is structured in the localization and extraction of primitives in each individual image, the characterization of the primitives by iconic and symbolic attributes, the correspondence analysis to match the primitives of the images and the reconstruction of a 3-dimensional object model on the basis of the matched primitives. The procedures can be performed straight forward, as well as being regarded as components of a recursive filter process for an incremental set-up of the object model. In both strategies automatically extracted data of each individual process are stored in a data and knowledge base, which is the central module to control the System. The processing steps can be assigned to a low or iconic level, a medium or symbolic level and a high or geometric level. In the iconic level, at first we use algorithms to detect and analyze image regions with gray-level discontinuities as well as shape from shading methods for image regions with homogenous grey-values. Depending on the grey value gradient, the entire image is described by point-primitives, line-primitives and region primitives. Each primitive is then characterized by specific attributes containing iconic information and additionally, in the symbolic level, by the relations of a primitive to its neighboring primitives. This information is used for matching primitives of image pairs and in finding corresponding primitives in ail acquired images. Image coordinates of the matched primitives obtained in the correspondence analysis enables reconstruction of the object model by photogrammetric standard procedures (e.g. bundle adjustment) and thus the change from symbolic to geometric level. The more images are analyzed, the more redundancy image information can be utilized to detect and reject primitive mismatches, increasing the reliability of the processes. In this dissertation the individual processing steps are commonly defined, such that this approach can be easily adapted to different applications by replacing functions by specially defined and more efficient functions for the actual application. Instead of point-primitives that are used as target primitives in this work, line-primitives and region primitives can be principally employed for reconstruction and parametric description of an object model. The main goal of this work is to give an overall approach from image acquisition to model reconstruction, utilizing redundant information for getting information about accuracy and reliability of the reconstructed model. The principle of using redundant information that can be automatically acquired is always recommended, if acquisition and definition of a priori information is difficult or impossible. This dissertation describes the principal method and gives an example, where processing steps are illustrated and efficiency and performance of the algorithms are discussed. Numéro de notice : 28091 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse étrangère Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=63438

Exemplaires(2)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
28091-01	33.30	Livre	Centre de documentation	Photogrammétrie - Lasergrammétrie	Disponible
28091-02	33.30	Livre	Centre de documentation	Photogrammétrie - Lasergrammétrie	Disponible