Drones Paired with Hyperspectral Imaging Paired with LiDAR to Locate Explosive Ordnance

Anti-personal/tank landmines, improvised explosive devices (IED), unexploded ordinances (UXO), and other abandoned explosive ordinances (EO) all pose long-lasting threats that are detrimental to areas of conflict. From 2015 to 2021, a total of 49,050 deaths/injuries were caused by EOs, and this number is only increasing. Current demining methods heavily rely on ground-based electromagnetic-induction (EMI); however, this method is costly, time consuming and puts personnel at risk. Recent advances in drone and remote sensing technology have allowed for the development of alternative remote methods to improve the efficiency in locating EOs. We used a Velodyne VLP-16 light detection and ranging (LiDAR) sensor attached to a DJI Matrice 600 drone platform to remotely identify EOs, specifically PFM-1 and VPMA-3 anti-personnel mines, TM-62M anti-vehicle mines, and 3 meter long 122 mm multibarrel rockets (MBRL). LiDAR data was acquired in dual return acquisition mode at 300 rpm and a flight speed of 1 m/s. Several of these EOs are being used in the current Russo-Ukrainian war, including: TM-62 anti-vehicle mines, PFM-1 landmines, and the MBRL rockets. Our LiDAR sensor was calibrated with a 18 m swath width to acquire 4630 points/m2  density and a 1.7 cm footprint resolution. The LiDAR data that was collected was post-processed to produce various derivative data such as: 3D point clouds, digital elevation models (DEM), digital surface models (DSMs), and derivative data products such as the total horizontal derivative (THD) filter. Processed data highlighted lateral spatial heterogeneity, which identified vertical and horizontal MBRLs, as well as surficial TM-62M anti-vehicle, TM62P anti-personnel mines and VPMA-3 landmines. PFM-1 landmines, the smallest of all EOs used, were not located, as the footprint resolution of the data collected was too small (1.7 cm) to clearly differentiate the ordinance from the environment. This pilot study allowed us to better understand the strengths and weaknesses of this method. We plan to further develop this technology by exploring the use of streamlined algorithms, applying alternative data processing workflows, and using sub-pixel techniques to improve the accuracy and efficiency of location. Refining data acquisition parameters, such as the speed and height of drone flight may also lead to further improvements in efficiency. In addition to location, a focus could also be placed on looking at intensity to identify material properties of EOs.