PETALS: a visual interface for landmine detection

Abstract

Post-conflict landmines have serious humanitarian repercussions: landmines cost lives, limbs and land. The primary method used to locate these buried devices relies on the inherently dangerous and difficult task of a human listening to audio feedback from a metal detector. Researchers have previously hypothesized that expert operators respond to these challenges by building mental patterns with metal detectors through the identification of object-dependent spatially distributed metallic fields. This paper presents the preliminary stages of a novel interface - Pattern Enhancement Tool for Assisting Landmine Sensing (PETALS) - that aims to assist with building and visualizing these patterns, rather than relying on memory alone. Simulated demining experiments show that the experimental interface decreases classification error from 23% to 5% and reduces localization error by 54%, demonstrating the potential for PETALS to improve novice deminer safety and efficiency.