Call number:
S 99.0139(384)
In:
Wissenschaftliche Arbeiten der Fachrichtung Geodäsie und Geoinformatik der Leibniz Universität Hannover, Nr. 384
Description / Table of Contents:
This dissertation shows the importance of using low-cost crowd-sourced information for the task of traffic regulator recognition (traffic signals, stop signs, priority signs, uncontrolled intersections), the cost of which in terms of time and money is much higher if standard technology is used for surveying. GPS trajectories can reveal the movement patterns of traffic participants, and the initial hypothesis that traffic regulations can be retrieved by mining the movement patterns imposed by traffic rules is verified. The predictive ability of the classifier becomes more accurate when static information derived from open maps
(OSM) is merged with dynamic features extracted from GPS trajectories. An extensive evaluation of the proposed methodology on three datasets, provided classification accuracy between 95% and 97%.
Type of Medium:
Series available for loan
Pages:
ix, 178 Seiten
,
Illustrationen, Diagramme, Karte
ISBN:
978-3-7696-5310-6
,
9783769653106
ISSN:
0174-1454
Series Statement:
Wissenschaftliche Arbeiten der Fachrichtung Geodäsie und Geoinformatik der Leibniz Universität Hannover Nr. 384
Language:
English
Note:
Dissertation, Gottfried Wilhelm Leibniz Universität Hannover, 2023
,
Contents
1. Introduction
1.1. From GPS tracks to Traffic Regulations
1.2. Motivation for Learning Intersection Traffic Regulations
1.3. Research Gap
1.4. Motivation for Learning Traffic Regulations from GPS Data
1.5. Research Objectives, Challenges and Contributions
1.6. Outline of the Thesis
1.7. Summary
1.8. Acknowledgements
2. Theoretical Background
2.1. Intersections and Intersection Traffic Regulations
2.1.1. Intersections
2.1.2. Intersection Traffic Regulations
2.2. Spatiotemporal Data and Movement Trajectories
2.2.1. The Global Positioning System
2.2.2. Sampling Frequency
2.2.3. The GPS Exchange Format: GPX
2.2.4. Movement Patterns in Spatiotemporal Data
2.2.5. Spatiotemporal Data Mining
2.2.6. Semantic Enrichment of Trajectories
2.2.7. Detecting Stop and Moves: the CB-SMoT Algorithm
2.3. Machine Learning
2.3.1. Machine Learning and Types of Learning
2.3.2. Supervised-Learning: Classification
2.3.3. Unsupervised-Learning: Clustering
2.3.4. Semi-supervised Learning
2.3.5. Active-Learning
2.3.6. Incremental Learning
2.4. Acknowledgements
3. Related Work
3.1. Existing Traffic Regulation Recognition Approaches
3.2. Static Categorization
3.2.1. Map-based Category
3.2.2. Image-based Category
3.3. Dynamic Categorization
3.3.1. Episode-based Category
3.3.2. Speed-profile Category
3.3.3. Movement-summarization Category
3.4. Hybrid-based Categorization
3.5. Discussion
3.6. Knowledge Gap
3.7. Acknowledgements
4. Traffic Regulation Recognition (TRR) from GPS Data
4.1. Introduction
4.2. Datasets
4.2.1. Dataset Requirements and Limitations
4.2.2. Datasets for Testing the Proposed Methods
4.2.3. Groundtruth Map Construction
4.3. Methodology
4.3.1. Detection of Stop and Deceleration Episodes
4.3.2. The Static Approach
4.3.3. The c-Dynamic Approach
4.3.4. The Dynamic Approach
4.3.5. The Hybrid Approach
4.3.6. Implementation and Classification Settings
4.4. Results
4.5. Discussion
4.6. Summary
4.7. Acknowledgements
5. TRR From GPS Data: One-Arm versus All-Arm Models
5.1. Introduction
5.2. Methodology
5.2.1. One-Arm vs. All-Arm Models
5.2.2. The Effect of Sampling Rate
5.2.3. Reduced Models
5.2.4. The Effect of Turning/No-Turning Trajectories
5.2.5. The Effect of Number of Trajectories
5.2.6. Application of Domain Knowledge Rules
5.2.7. Classification Settings
5.3. Results
5.3.1. One-arm vs. All-arm Models
5.3.2. Testing the Effect of Sampling Rate
5.3.3. Reduced Models
5.3.4. Testing the Effect of Turning Trajectories and Examining an Optimal Number of Trajectories
5.3.5. Testing the Effect of the Number of Trajectories on Classification Performance
5.3.6. Misclassification Analysis
5.3.7. Applying Domain Knowledge Rules
5.4. Discussion
5.5. Summary
5.6. Acknowledgements
6. TRR with Sparsely Labeled and Stream Data
6.1. Introduction
6.2. TRR with Clustering
6.3. TRR with Self-Training, Active Learning and Cluster-then-Label
6.3.1. TRR with Self-Training: Using Labeled and Unlabeled Data
6.3.2. TRR with Active Learning
6.3.3. TRR with the Cluster-then-Label Algorithm
6.3.4. Comparison of All Tested Methods
6.4. Learning Transferability: Training on City A and Predicting on City B
6.5. Incremental (Online) Learning
6.6. Summary
7. Conclusions and Outlook
7.1. Research Questions Addressed in this Thesis
7.2. Outlook
List of Acronyms
Index
A. Appendix
List of Figures
List of Tables
Bibliography
Curriculum Vitae
Acknowledgements
,
Sprache der Kurzfassungen: Englisch, Deutsch
Location:
Lower compact magazine
Branch Library:
GFZ Library
