Centre de documentation
scientifique

Robust estimators for geodetic transformations and GIS / Entela Kanani (2000)  

Public Titre : Robust estimators for geodetic transformations and GIS Type de document : Thèse/HDR Auteurs : Entela Kanani, Auteur Editeur : Zurich : Institut für Geodäsie und Photogrammetrie IGP - ETH Année de publication : 2000 Collection : IGP Mitteilungen, ISSN 0252-9335 num. 070 Importance : 133 p. Format : 21 x 30 cm ISBN/ISSN/EAN : 978-3-906467-26-9 Note générale : Bibliographie non accessible dans la collection Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Statistiques [Termes IGN] compensation par modèles [Termes IGN] estimateur [Termes IGN] méthode des moindres carrés [Termes IGN] méthode robuste [Termes IGN] simplification de contour [Termes IGN] transformation de coordonnées [Termes IGN] transformation de Helmert [Termes IGN] valeur aberrante [Termes IGN] vectorisation Résumé : (Auteur) Most parameters in geodesy are determined by statistical estimator procedures. The method most commonly employed is that of Least Squares - an optimal estimator for normal distributed measurements. In practical evaluation, however, there often occur many observations which do not conform to the expected normal distribution. This occurs, for example, when a part of the observations contain outliers. With the help of robust statistics, it is possible to solve such problems, since they are not sensitive to outliers. This work concerns the evaluation of data using robust methods, particularly in the fields of linear transformation (geodesy) and vectorisation of areal objects from scanned topographic maps for Geographic Information Systems (GIS). Both the theoretical basis and the practical implementation are described. Besides the classical method of data evaluation, the Least Square, it is now possible to determine the transformation parameters at the geodetic transformation, by two robust estimators: the LMS estimator (from the class of the high breakdown point estimators) and the BIBER estimator (from the class of bounded influence estimators). It is shown that the LMS estimator, as a very good outlier detector, can be used when data is highly contaminated (up to 50%); whereas the BIBER estimator is more appropriate in instances where the number of outliers is small, and the assumption of the normal distribution can still be accepted. A combination of both estimators gives optimal results, even when the data are contaminated (up to 50%). A novel application in the field of GIS and cartography is demonstrated, involving the employment of robust techniques for the automatic vectorisation of buildings extracted from topographic and general maps. This application was motivated by the growing need for the automatisation of data acquisition processes in GIS, which are otherwise expensive, and laborious. An appropriate adjustment model, which incorporates robust techniques (BIBER estirnator) for the evaluation of data, delivers excellent results, in a short period of time, using a fully automatic process. The developed algorithms are implemented in prototype software and have shown to be good tools for the evaluation of data in these two fields. Note de contenu : 1. ESTIMATION OF PARAMETERS, IN GEODESY - The Adjustment Model in Geodesy - The Robust Adjustment Model in Geodesy 2. CLASSES OF ROBUST METHODS - The Huber-Type Estimator - Bounded Influence Estimates - High Breakdown Point Estimates 3. ROBUST ESTIMATORS FOR GEODETIC-TRANSFORMATIONS - Linear Transformations - Introduction - Linear Transformations in Geodesy - Determination of the Unknown Parameters by Least Square Method - Determination of the Unknown Parameters by Robust Estimators - Applications and Results 4. ROBUST ESTIMATORS IN THE AUTOMATICS STRUCTURING OF RASTER DATA - Topographic Map - From Analogue Copy to Numerical Data - Vectorisation of Topographic Features - Pattern Recognition based Vectorisation - Vectorisation - Contour Extraction - Simplification and Smoothing of Digitised Lines - Adjustment of the Approximated Contour - New Approach Numéro de notice : 11429 Affiliation des auteurs : non IGN Thématique : MATHEMATIQUE Nature : Thèse étrangère Organisme de stage : Swiss Federal Institute of Technology Zurich En ligne : https://ethz.ch/content/dam/ethz/special-interest/baug/igp/igp-dam/documents/PhD [...] Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=54403