Fontana, Marco 
ORCID: 0000-0003-0703-6535
(2024)
Multi-target detection and tracking using Distributed Acoustic Sensing.
PhD thesis, University of Liverpool.
|
Text
201445987_Jan2025.pdf - Author Accepted Manuscript Download (17MB) | Preview |
Abstract
Distributed Acoustic Sensing (DAS) is a groundbreaking technology that uses fibre optic cables to detect and monitor acoustic signals along their entire length. Unlike traditional localised point sensors, DAS transforms a standard fibre optic cable into a continuous, highly sensitive array of acoustic sensors, enabling comprehensive monitoring over extensive distances. This innovative approach has revolutionised various fields by providing real-time data acquisition with high spatial resolution and sensitivity. DAS operates by transmitting laser pulses through an optical fibre and analysing the resulting backscattered light. Variations in the backscattered light, caused by vibrations generated by vehicle and train movements, can be detected and processed to provide detailed information about traffic flow and patterns. To achieve this, raw data must be processed in real time by a detector that can locate acoustic events along the fibre. Furthermore, a tracker must use the noisy measurements from the detector to estimate the locations of the targets of interest, considering clutter and misdetection events. The focus of this thesis is on developing and optimising methodologies for accurately detecting and tracking multiple targets using DAS, with the aim of providing the theoretical basis for designing a real-time traffic monitoring system based on DAS data. Chapter 3 introduces an efficient and scalable approximation of the Poisson multi-Bernoulli Mixture (PMBM) filter. This approach involves defining a measurement-driven clustering algorithm to break down the data association problem into several subproblems. This contribution is essential for effectively tracking a large number of vehicles in real-world traffic scenarios. Chapter 4 introduces an extended target tracking approach designed for tracking trains using DAS data. Building upon the clustered version of the PMBM filter proposed in Chapter 3, the proposed method provides a solution to handle asymmetric noise within the set of measurements returned by each target at each time step, as well as merged measurements resulting from occlusion. When processing DAS data, the presence of a vehicle cannot be based solely on a single point of time, due to the noise generated by external sources and suboptimal coupling between the fibre and the road surface. Chapter 5 presents a method to detect vehicle trajectories in short time windows using the notch periodogram. This method iteratively estimates trajectory segments and notches their contribution in the original data, providing remarkable detection performance in high-density traffic scenarios. The proposed detector returns sets of trajectory measurements, which are then input into a specially designed tracker presented in Chapter 6. In this tracker, each trajectory measurement is a straight line segment defined by two locations: one at the beginning and one at the end of the time window.
| Item Type: | Thesis (PhD) |
|---|---|
| Uncontrolled Keywords: | Bayesian filtering, random finite sets, multi-target tracking, Poisson multi-Bernoulli mixtures, vehicle detector, notch periodogram, distributed acoustic sensing |
| Divisions: | Faculty of Science and Engineering Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 19 Aug 2025 15:25 |
| Last Modified: | 06 Oct 2025 13:10 |
| DOI: | 10.17638/03189781 |
| Supervisors: |
|
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3189781 |
Altmetric
Altmetric