Call number:
SR 99.0038(818)
Description / Table of Contents:
The improvement of measurement and particularly surveying technologies results in a large as well as rapidly increasing amount of spatial data. These data stem from various measurement techniques as well as platforms and, therefore, may compile quite different densities, qualities, and error characteristics. Effective tools are required to understand and interpret them. The challenges include efficient processing, robustness against data flows and uncertainty, rationality of modeling, and the potential of automation and learning. This thesis presents an exploration of the use of statistical models and related techniques in spatial data analysis. The foundation of the methodology employed in the scope of this thesis consists of Bayesian statistics and Markov models. Selected approaches conceived by the author, including 3D building reconstruction, semantic building classification, pattern recognition in trajectories, and segmentation of RGBD data, demonstrate their potential in spatial data modeling and interpretation.
Type of Medium:
Series available for loan
Pages:
88 Seiten
,
Illustrationen, Diagramme, Karten
ISBN:
978-3-7696-5229-1
,
9783769652291
ISSN:
0065-5325
Series Statement:
Deutsche Geodätische Kommission bei der Bayerischen Akademie der Wissenschaften : Reihe C, Dissertationen 818
URL:
https://publikationen.badw.de/de/045060722
Classification:
Photogrammetry, Remote Sensing
Language:
English
Note:
Habilitationsschrift, Universität Gottfried Wilhelm Leibniz Universität Hannover, 2018
,
Contents
I Synopsis
1 Introduction
1.1 Spatial data and the challenges
1.1.1 Characteristics of spatial data
1.1.2 Challenges
1.2 Statistical models
1.2.1 Bayesian statistics
1.2.2 Markov models
1.3 Scope and organization
2 Building reconstruction
2.1 Problem statement
2.2 Data – LiDAR and imagery
2.3 Model – generative models for buildings
2.3.1 Primitive-based modeling
2.3.2 Extension for building generalization
2.4 Reversible Jump Markov Chain Monte Carlo
2.5 Model selection
2.5.1 Bayesian model selection
2.5.2 Information entropy and model size estimation
2.6 Related work
2.7 Conclusion and Remarks
3 Building classification
3.1 Problem statement
3.2 Data – building footprints
3.3 Model – Markov Random Field for building network
3.3.1 Markov Random Field
3.3.2 Network of buildings
3.3.3 Local geometric features
3.3.4 Contextual relationship
3.4 Gibbs sampler
3.4.1 Metropolis-Hastings Algorithm
3.4.2 Gibbs sampling
3.5 Related work
3.6 Conclusion and remarks
4 Anomaly detection in trajectories
4.1 Problem statement
4.2 Data – GPS trajectories
4.3 Model – Hidden Markov Model for trajectory
4.3.1 Hidden Markov Model with dynamic orders
4.3.2 Long-term spatial and temporal features
4.4 Bayesian filter for belief inference
4.4.1 A dynamic Bayesian filter
4.4.2 Belief inference
4.4.3 Collective behaviors
4.5 Related work
4.6 Conclusion and remarks
5 RGBD Segmentation
5.1 Problem statement
5.2 Data – RGBD
5.3 Model – A novel synthetic model for spatial data parsing
5.3.1 Synthetic volume primitives – SVP
5.3.2 Freeform object voting
5.4 Global optimization with Markov Random Field
5.5 Related work
5.6 Conclusion and remarks
6 Conclusion and Discussion
6.1 Answers to Challenges
6.2 A Start of the Exploration: Limits and Potential
6.2.1 Characteristics of Big Data
6.2.2 Balance between Top-down and Bottom-up
Bibliography
II Publications
Publication list
Not included publications
,
Deutsche und englische Zusammenfassung
Location:
Lower compact magazine
Branch Library:
GFZ Library
Permalink
