AI-Assisted Landmine Contamination Prediction — NEC, JICA and Cambodia's CMAC Pilot Model Unsurveyed Minefields

Cambodia carries one of the world's heaviest landmine and unexploded-ordnance legacies; as of early 2024 an estimated 1,576 km² of land remained contaminated (including 348 km² by anti-personnel mines), with 19,834 people killed and 45,252 injured by mines/UXO between 1979 and 2024. The Cambodian Mine Action Centre (CMAC), the national demining authority, cleared about 200 km² and destroyed 99,874 mines/UXO items in 2024, working toward Cambodia's Ottawa Treaty target of being mine-free by 2030.

NEC Corporation and NEC Laboratories Europe partnered with JICA and CMAC on a proof-of-concept AI model to speed up the costly manual 'non-technical survey' process used to identify which unsurveyed areas warrant clearance.

The model analyses CMAC's historical landmine-location records, resident-reported hazard information and open geospatial data (including terrain and infrastructure layers) to predict which unsurveyed areas are most likely to contain landmines.

In a demonstration area of roughly 1 million m², the model's predictions matched more than 90% of actually known landmine locations; the project won GICHD's 2025 Innovation Award for 'Improving the Accuracy of Estimating Explosive Ordnance Hazardous Areas,' presented in Luxembourg in October 2025 to more than 400 mine-action professionals.

The result is explicitly a proof-of-concept: the roughly 1 km² demonstration area is a tiny fraction of Cambodia's contaminated footprint of over 1,500 km², and no public source specifies a funding commitment or timeline for scaling the model to nationwide survey planning.