Detection of Seismic and Acoustic Signals With Serial Network Data Fusion: Demonstration Against Atmospheric Explosions
Description:
Explosion monitoring conventionally operates in siloes: seismic, infrasonic, and other ground or space based sensors process individual modalities to perform functions like detection. Such siloed detection excludes evidence collected by other modalities that may be leveraged to achieve greater detection rates, lower nuisance alarms, and lower thresholds. Work conducted at Los Alamos National Laboratory seeks to increase detection performance by removing these siloes to facilitate data fusion.
We thereby present a proto-type system to fuse multiple explosion-sourced (“target” sourced) modalities into a serial architecture that shares detector output and makes collective (system-wide) decisions regarding source identity. The system shares this data by coupling the thresholds of its constituent detectors. Decisions made by one detector then inform a downstream detector, which processes a second modality. This downstream detector then adjusts its threshold according to the decision made by the upstream detector. Our system selects the coupling strength between these thresholds to achieve a maximum system-wide true positive rate and maintain an acceptable, system-wide false positive rate. The resulting system-wide detector achieves a superior performance relative to that of the best, constituent detector. We present these performance characteristics on a real data recorded from an explosion sourced on the US East Coast.
Our system thereby exploits the theory of distributed detection (Viswanathan et al., 1988), which has been exploited in multiple sensor networks. It remains under-leveraged in explosion monitoring. If successfully demonstrated in explosion monitoring scenarios, we argue that such systems can provide a non-incremental gain to monitoring capability that mimics the teamwork that of humans at the analyst-review stage of processing.
Session: Advancements in Forensic Seismology and Explosion Monitoring - III
Type: Oral
Date: 5/2/2024
Presentation Time: 03:00 PM (local time)
Presenting Author: Joshua
Student Presenter: No
Invited Presentation:
Authors
Joshua Carmichael Presenting Author Corresponding Author joshuac@lanl.gov Los Alamos National Laboratory |
Richard Alfaro-Diaz rad@lanl.gov Los Alamos National Laboratory |
Tess Light tlavezzi@lanl.gov Los Alamos National Laboratory |
Philip Blom pblom@lanl.gov Los Alamos National Laboratory |
Christine Gammans cgammans@lanl.gov Los Alamos National Laboratory |
Brent Delbridge delbridge@lanl.gov Los Alamos National Laboratory |
Chris Carr cgcarr@lanl.gov Los Alamos National Laboratory |
Michael Begnaud mbegnaud@lanl.gov Los Alamos National Laboratory |
|
Detection of Seismic and Acoustic Signals With Serial Network Data Fusion: Demonstration Against Atmospheric Explosions
Category
Advancements in Forensic Seismology and Explosion Monitoring