AI-Powered Drones Now Capable of Detecting Plastic Landmines in Ukraine

A Fresh Approach to Spotting Plastic Anti-Personnel Mines

According to НВ — Техно: Researchers at Binghamton University have created a novel detection method for plastic anti-personnel mines, combining a drone, a camera, and the YOLO machine learning algorithm. Successful tests were conducted using inert PFM-1 mines-widely deployed Soviet-era anti-personnel devices-alongside 3D-printed replicas. Field operations require only a lightweight laptop, a drone, and a camera, with no internet connection needed.

Modern anti-personnel mines feature plastic casings, making them notoriously difficult to find. Standard metal detectors fail to register these plastic shells, and geophysical techniques like ground-penetrating radar, magnetometry, and electromagnetic induction prove less effective. In active combat zones such as Ukraine, scattered mines often lie close to the surface, while in post-conflict areas they may become buried or hidden within the landscape.

Training the Algorithm

The team trained the YOLO algorithm using inert PFM-1 mines and their replicas, placed in various locations within Binghamton University's natural preserve. Two distinct YOLO models were developed: the first trained exclusively on PFM-1 mines, and the second programmed to identify both PFM-1 mines and random objects from a standard dataset. However, the second model yielded lower performance metrics. The algorithm's training phase lasts anywhere from a few hours to a full day.

This is a first-pass analysis to determine if the terrain is a potentially suspicious hazardous area.

- Sharifa Karwandyar, graduate of Binghamton University's Geology Department

According to Alex Nikulin, an associate professor in Binghamton University's Department of Earth Sciences, 'Caring for a wounded soldier is harder than caring for a dead one. They are designed to wound, not kill. They are intentionally built for that purpose, and their entire construction is meant to be invisible.' This new technique opens up safer possibilities for detecting and neutralizing dangerous mines in conflict zones.

In 2024, Mateo Dulce Rubio, a graduate student at Carnegie Mellon University, also introduced a mine detection system called RELand, which leverages artificial intelligence, machine learning, geographic information, and socio-demographic data. These innovations in mine detection technology could significantly enhance safety in conflict-affected regions.

The development of this new method for detecting plastic anti-personnel mines marks a critical step in addressing the dangers posed by leftover battlefield ordnance. Where traditional detection methods fall short, emerging technologies-especially drones and machine learning algorithms-can substantially reduce risks for both civilians and military personnel. It also underscores the vital role of innovation in security, particularly in areas scarred by armed conflict.

As advancements in drone technology continue, the integration of AI in detecting explosives is becoming increasingly critical. For instance, a recent innovation from the US allows a device to identify explosives from an impressive height of 200 feet. This development complements the efforts in Ukraine, where detecting plastic landmines remains a pressing challenge. To explore this groundbreaking technology and its implications, read more about how it enhances aerial surveillance capabilities in this related article.