top of page
Copy of GabeJasper2.jpg

We love science and making the world a safer place. Scroll down to see the projects our team is working on to help detect landmines and UXO.

Leveraging what we believe to be the most robust and extensive drone-based imagery dataset of surface munitions, we are training a convolutional neural network (CNN) to detect and categorize 10 different types of landmines and UXO. Categorizations will include projectiles, anti-personnel landmines, anti-vehicle landmines, 40mm grenades and cluster munitions. This method is designed to work for surface munitions on any environment with munitions ranging from fully visible to heavily obscured. Initial tests yield over 85% accuracy averaged across all objects. We are currently hard at work on an offline application that can be deployed in the field by deminers across the world.


The PFM-1 is a plastic, scatterable landmine that has been used in conflicts for over 40 years and has resulted in immense suffering for its many victims. Since 2016 we have been creating a method to automatically detect these mines with the potential to use the technology to detect a wide range of other landmines and UXO. Since the mines are deployed from mortars, helicopters and airplanes and land on the surface of the ground, we can use commercial drones that collect visual and thermal images paired with cutting-edge machine learning techniques to detect the mines in many different environments and return their exact coordinates, making demining safer and more efficient.

Unexploded ordnance (UXO) are explosive weapons that failed to explode upon deployment and often remain in or on the ground for decades after a conflict, posing a constant risk of explosion. The BM-21 is a widely-used rocket launcher whose rockets often do not explode upon impact but instead years later, leading to the injury and death of its victims and contaminating environments with hazardous metals. The large size and high metallic content of these rockets gave us the idea to create an aerial-magnetic drone system to detect them. Results of our field trials prove this method to be significantly cheaper and faster than current methods for finding these rockets.


Snakebot, the demining robot


A large portion of mine clearance budget and time is spent on clearing brush and vegetation. Current mine clearance methods relying on metal detection are time and labor intensive, have a high false-positive rate, and present considerable operator risk to explosive ordnance disposal (EOD) personnel. There are currently no automated methods to detect and accurately locate UXOs in heavily vegetated areas without prior vegetation clearance. To combat this issue, we built a remote-controlled ground-coupled “snakebot” that can navigate the difficult terrain on the ground, equipped with a Geometrics MFAM magnetometer. The snakebot has been field tested in real world minefields in Cambodia, and is currently in the prototype, proof-of-concept stage.

The MineMarker™

The MineMarker is a drone attachment that dispenses brightly colored chips that mark the physical location of mines and UXO from above. Marking the location of an explosive hazard prior to demining clearance efforts improves the situational awareness and safety of our demining personnel. The MineMarker works in conjunction with our machine learning model the DRCNN, which first detects suspected locations of the mines from a drone survey. After the mines are detected using the DRCNN, the MineMarker physical marks and visually confirms each of the detections, all without a person ever stepping foot into the minefield. 

bottom of page