Researchers Develop Multisensor Drone Mine Dataset

Researchers at Rochester Institute of Technology and collaborators have developed a large, georeferenced multisensor dataset and evaluated drone-based sensors for land mine and unexploded ordnance detection. The collection covers over 140 inert targets in Oklahoma and 110 PFM-1 replicas in Belgium, captured with hyperspectral, multispectral, thermal, RGB, LiDAR, SAR, GPR, electromagnetic-induction, and magnetometer sensors. They also propose uncertainty-aware AI methods to improve detection reliability for demining operations.