Activity Number:
|
661
- Statistical Approaches to High-Dimensional Modeling and Real-World Problems
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 3, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
ENAR
|
Abstract #323652
|
View Presentation
|
Title:
|
Sensor Fusion Approach for Detecting Landmines
|
Author(s):
|
Andrew Cho* and Colin Rundel
|
Companies:
|
Duke University and Duke University Statistical Science
|
Keywords:
|
Spatial ;
Thin plate spline ;
Fusion
|
Abstract:
|
Electromagnetic induction (EMI) and ground penetrating radar (GPR) are two sensors types that are commonly used to detect landmines. While EMI sensors fare well in detecting the existence of metal content in the ground, this alone cannot discriminate a landmine from harmless scrap metal or detect an explosive device made of plastic. Similarly, GPR sensors are useful for detecting disturbances / changes in the density of soil but also highly sensitive to changes in surface conditions. One practical challenge of mine clearing is the need to correctly and efficiently identify the devices over a wide area. This paper considers a statistical identification algorithm that combines signal data from both EMI and GPR sensors to create a probabilistic spatial model of a landmine occurrence across a scanned region. Due to the scale of the data we employ independent thin plate spline models of each sensor to generate a smooth latent surfaces. These surfaces are then fused to create a probability unified surface which is the basis for identifying and flagging target locations which are predicted to contain landmines.
|
Authors who are presenting talks have a * after their name.