Drones and AI in Modern Plastic Landmine Detection

Humanitarian demining teams are revolutionizing the clearance of former conflict zones by pairing drone-mounted sensors with artificial intelligence to locate buried plastic antipersonnel landmines. Traditional metal detectors often struggle to find these munitions due to their minimal metal content, making manual clearance slow and hazardous. By utilizing thermal inertia imaging to detect subsurface temperature anomalies and multispectral cameras to identify soil and vegetation disturbances, unmanned aerial vehicles (UAVs) can safely scan vast areas from above. Deep learning algorithms then process this complex aerial data to pinpoint the exact coordinates of hidden explosives, drastically accelerating the demining process and saving lives.

Buried plastic landmines present a severe challenge because they cannot be reliably found using standard electromagnetic induction. To overcome this, drones are equipped with high-resolution thermal cameras that leverage the principle of thermal inertia. Plastic mines and the disturbed soil above them absorb and release solar heat at a different rate than the surrounding undisturbed earth. Drones fly over hazardous areas during peak thermal transition periods—typically just after sunrise and sunset. As the ground heats up or cools down, the plastic casing of a buried mine acts as an insulator, creating a distinct temperature anomaly on the surface directly above it. The thermal sensors capture these subtle variations, which are invisible to the naked eye.

Alongside thermal imaging, drones utilize multispectral sensors to capture data across various wavelengths of light, including visible, near-infrared, and short-wave infrared bands. This technology identifies physical and chemical changes in the environment caused by the presence of a landmine. For example, the initial burial of a mine alters the soil's compactness and moisture retention, which shows up in specific spectral bands. Furthermore, over time, explosive chemicals such as TNT can leach out of the plastic casing and into the surrounding soil. This contamination induces subtle stress in the overlying vegetation, altering the plants' chlorophyll levels and light reflectance—a phenomenon that multispectral sensors easily detect.

Collecting thermal and multispectral data generates thousands of high-resolution images filled with environmental noise, as rocks, animal burrows, and roots can mimic the signatures of landmines. This is where artificial intelligence becomes essential. Demining teams feed the overlapping thermal and multispectral datasets into convolutional neural networks (CNNs) trained specifically on the signatures of known ordnance. The AI fuses the data layers together, cross-referencing a thermal anomaly with the corresponding multispectral soil and vegetation data. By analyzing the shape, size, thermal gradient, and spectral reflectance simultaneously, the AI filters out false positives and identifies buried plastic mines with high precision.

Once the AI finishes processing the aerial surveys, it generates a high-accuracy, georeferenced map using Real-Time Kinematic (RTK) GPS data collected by the drone. This map highlights the exact coordinates of suspected plastic landmines within a few centimeters. Ground-based demining personnel use these maps to navigate directly to the confirmed threats, bypassing large swaths of safe land that would otherwise require slow, manual prodding. By transforming a blind search into a targeted removal operation, the combination of drone-mounted imaging and AI reduces the time, cost, and human risk required to return safe land to local communities.