Enhanced Tunnel Detection with Multiple Sensor Technologies

Project Overview

Detection and characterization of tunnels and subsurface structures is a difficult problem in border protection and nonproliferation. No one sensor technology has been shown effective for detecting more than a small subset of tunnels. This study explored the feasibility of using a bi-directional stochastic anomaly exposure (SAE) peak filtering approach to detect tunnels by fusing geographical information system (GIS) data layers with data layers of simulated sensor detections along a portion of the southern border. The GIS data layers contain locations of places (such as buildings or vegetation) where tunnels could be hidden. Data layers for locations and other attributes of possible sensor detections along the southern border are derived from sensor readings for magnetometers, gravity gradiometers and low frequency ground penetrating radars that are simulated with existing codes. Based on these simulations, models for tunnel detection are designed for each of these sensors individually, for each combination of two sensors and for the combination of all three sensors. Each model is matched to the GIS and sensor data layers. Tunnel detection effectiveness for the individual sensors and the various combinations of sensors is assessed by comparing the detection results to ground truth. The statistical calculations demonstrate that the combination of three sensors performs at least as well as combinations of two sensors which performs at least as well as individual sensors. They show the approach is computationally practical, and one can improve tunnel detection performance by combining data from multiple sensors. The next step is for the Feasibility Study (FS) to be expanded to incorporate higher resolution data, more rigorous codes and simulations, and real sensor data to apply this data fusion methodology for tunnel detection in a real-world environment.

Mission Impact

Methods of sensor data fusion for detecting tunnels advance the Earth and Atmospheric Core Competentecy and Defense and Homeland Security mission area, specifically advancing our national security capabilities in border protection.

Publications, Presentations, and Patents

[1] D. Paglieroni, "Detection of Anomalous States in Sensor Data", IL-13683, U.S. Patent Pending, Filed July 2022.

D. Paglieroni, D. Chambers, W. Kirkendall, and E. Raber. "Sensor Suitability Scoring Methodology for Tunnel Detection." Lawrence Livermore National Laboratory, Livermore, CA. Technical Report LLNL-TR-831511, May 30, 2023.