Outlier Detection of Poisson-Distributed Targets Using a Seabed Sensor Network

Conference Information

Event: IEEE OCEANS 2025 – Great Lakes
Website: https://greatlakes25.oceansconference.org/
Dates: September 29 – October 2, 2025
Location: Navy Pier, Chicago, Illinois

I presented my paper “Outlier Detection of Poisson-Distributed Targets Using a Seabed Sensor Network.”

Paper & Slides

• 📄 Paper (PDF): Outlier Detection of Poisson-Distributed Targets Using a Seabed Sensor Network
  :contentReference[oaicite:0]{index=0}

• 📊 Presentation Slides (PPTX):
  :contentReference[oaicite:1]{index=1}

Abstract

This work presents a framework for identifying and detecting spatial commission outliers in maritime environments using seabed acoustic sensor networks and log-Gaussian Cox process (LGCP) models.
The target arrival process is modeled as a mixture of normal and outlier processes, and a second-order approximation of the outlier-arrival probability is derived to incorporate both the mean and variance of the normal intensity function. This yields more accurate classification of atypical events compared to mean-only methods.

To enhance detection capability, we integrate a real-time, near-optimal sensor placement algorithm that adjusts sensor locations as the estimated outlier intensity evolves. Using real ship traffic data near Norfolk, Virginia, the proposed framework demonstrates improved classification performance and higher detection rates for outlier targets in simulation.

Context of the Work

This presentation builds upon my previous research on:

• Log-Gaussian Cox process modeling for uncertain target arrivals
• Real-time sensor placement and void probability maximization
• Approximation analysis using Jensen’s inequality
• Application of INLA for Bayesian spatial modeling

This conference version focuses on the outlier detection extension and the integration of closed-form approximations for real-time maritime surveillance.

Citation

M. Kim, D. J. Stilwell, and J. Jimenez,
“Outlier Detection of Poisson-Distributed Targets Using a Seabed Sensor Network,”
presented at IEEE OCEANS 2025, Chicago, IL.