I-Corps: Delivering Integrated Technology for Automated Wide-Area Humanitarian Mine Action Surveys

The broader impact/commercial potential of this I-Corps project is the development of technologies that effectively detect and clear landmines over wide surface areas. Humanitarian demining is the process where agencies (governmental or otherwise) survey a potentially hazardous area to detect, identify, and subsequently clear any explosive hazards (landmines or unexploded ordinance) in a post-conflict region. The goal of humanitarian demining is to release previously explosive-contaminated land back to impacted communities to allow the land to be used and developed without fear of injury and death caused by landmines and unexploded ordnance. The enormous variety of explosives, environments in which they lie, and conditions in which they may be found, create a multi-faceted problem that requires innovative technological solutions. This project is a demonstration of the powerful and wide-reaching capabilities of drones and machine learning in humanitarian mine action applications. The implementation of this technology will dramatically increase the rate and safety of demining, increasing the safety of demining operators, and allowing communities to reclaim previously inaccessible land for future development. This I-Corps project is based on the development of an integrated artificial intelligence-assisted unmanned aerial vehicle drone survey platform that a demining agency could quickly and effectively use to perform initial wide-area surveys. This technology will save money and time on subsequent demining activities in areas of critical contamination. The methods can be expanded to detect not only direct evidence of landmines and unexploded ordinances, but also indirect evidence such as explosion craters, minefield warning signs, or dead livestock potentially killed by explosives. Combining state-of-the-art deep learning technology, commercially available drones, and miniaturized optical sensors, this team created new solutions. This technology involves drone survey methods combined with innovative deep learning algorithms. The first step consists of integrating commercial visual light, thermal and multispectral drones to survey areas that are suspected to be contaminated with explosives. After the survey is complete, the deep learning method is used to generate the predicted coordinate locations of the explosives. In addition, a high-quality map of the suspected hazardous region is produced with each predicted explosive marked and labeled with its predicted class. The entire process takes place without using the internet, is user friendly, and can detect and classify most of the aerially-visible landmines and unexploded ordinances at a rate of 800 square meters per hour. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.