Ακριβής και κλιμακούμενη ανακατασκευή διαδρομών GPS με χαμηλό ρυθμό δειγματοληψίας
View/
Keywords
Map‐matching ; GPS trajectories ; Trajectory reconstruction ; Interpolation ; Apache Kafka ; Streaming ; Scalability ; GPS ; Vehicle data ; GPS data ; Map‐match ; Map matchingAbstract
This thesis presents an accurate and scalable pipeline for reconstructing vehicle trajectories from low‐sampling‐rate GPS data. We introduce three complementary algorithms—Trajectory Refinement, Curve Interpolation, and Trajectory Combination—that operate as lightweight post‐processing atop HMM‐based Map‐Matching. The methods correct road‐segment transitions, replace unrealistic straight‐line gaps with network‐constrained curves, and, where applicable, exploit historical trips on the same route to fill in missing segments. Beyond the algorithms, we design a streaming architecture on Apache Kafka (with Docker‐based deployment) that ingests, processes, and persists trajectories at scale via partitioned producers/consumers. We evaluate on 20 trajectories against manually verified ground truth generated with pgMapMatching. Compared to a baseline HMM Map‐Matcher, our approach improves overlap and classification metrics (e.g., F1 from 0.37 to 0.70, precision ≈0.71, recall ≈0.69), reduces route mismatch fraction (about 35% relative reduction) and alignment error (about 48% reduction), and yields more realistic path length (length index ≈1.07). Kafka experiments on up to 1,000 trajectories and 300K messages show near‐linear speedups up to 3–4 consumers (peaking around 4.4× with five) with diminishing returns beyond that point.
Postgraduate Studies Programme
Πληροφοριακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
38Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα