2025_programme: Hierarchical Multi-Target Tracking in Underwater Environments

• Day: June 17, Tuesday
    Location / Time: B. ERATO at 17:20-17:40
• Last minutes changes: -
• Session: 28. Signal and Image Processing
  Organiser(s): N/A
  Chairperson(s): Gabriele Morra
• Lecture: Hierarchical Multi-Target Tracking in Underwater Environments
  Paper ID: 2304
  Author(s): Joshua Wakefield, Stewart Haslinger, Jason Ralph
  Presenter: Joshua Wakefield
  Abstract: Passive sonar tracking is essential for various underwater applications, including defence, surveillance, and marine life monitoring. Unlike active sonar, passive systems rely solely on received signals, making tracking particularly challenging in the ocean environment due to high noise levels, multipath propagation, and limited measurement information. Existing approaches typically focus on applying a single filtering method tailored to a specific scenario. However, as ocean environments and signal conditions vary, no single filter is consistently optimal, leading to degraded tracking performance in changing conditions. This may be offset by utilising a more computationally intensive solution, but at the cost of increased computational complexity, making real-time tracking impractical in resource-constrained systems. A universal tracking filter is often inefficient in this setting, as different filters perform optimally under different conditions. Selecting the most suitable filter requires a principled approach that accounts for varying acoustic signal-to-noise ratios (SNR) and measurement uncertainties present in the underwater environment. This paper proposes an adaptive filtering framework that uses SNR as a decision boundary to select the most appropriate tracking filter dynamically. The choice of filter depends on the SNR, as high SNR conditions favour linear Gaussian approximations, while lower SNR scenarios benefit from more flexible, non-linear and non-Gaussian Sequential Monte Carlo (SMC) methods. By exploiting the available signal information, the framework adapts to changing measurement conditions, ensuring that the selected filter aligns with the reliability and uncertainty of the signal. Simulation results demonstrate the effectiveness of the proposed approach in passive sonar tracking scenarios. The framework adaptively selects between a linear Kalman Filter (KF), and a SMC filter based on the SNR, leading to improved computational efficiency while maintaining robust tracking performance. The results highlight the conditions under which each filter performs best, showcasing the benefits of an adaptive selection strategy over a fixed-filter approach.
    Download the full paper
  This paper is a candidate for the "Prof. Leif Bjørnø Best Student Paper Award (for students under 35)"
• Corresponding author: Mr Joshua Wakefield
  Affiliation: University of Liverpool
  Country: United Kingdom