The detection of explosive ordnance (EO) objects is experiencing a period of innovation driven by the convergence of new technologies including artificial intelligence (AI) and machine learning, open-source intelligence (OSINT) processing, and remote mobility capabilities such as drones and robotics.1 Advances are being made on at least two tracks: in the automated searching of photographic image archives, and in the real-time detection of objects in the field.2 Different technologies are responsive to different types of EO detection challenges, such as objects that are buried, semi-buried, or partially damaged. Computer vision—a type of artificial intelligence (AI) that enables computers and systems to derive meaningful information from digital images, videos, and other visual inputs, and take actions or make recommendations based on that information—is a promising AI technology that can greatly enhance humanitarian mine action (HMA), as well as support evidentiary documentation of the use of EO that are prohibited under international humanitarian law. This article describes a computer vision algorithm creation workflow developed to automate the detection of the 9N235/9N210 cluster submunition, a heavily deployed munition in the Ukraine conflict. The six-step process described here incorporates photography, photogrammetry, 3D-rendering, 3D-printing, and deep convolutional neural networks.3 The resulting high-performance detector can be deployed for searching and filtering images generated as part of OSINT investigations and soon, for real-time field detection objectives.
Harvey, Adam and LeBrun, Emile
"Computer Vision Detection of Explosive Ordnance: A High-Performance 9N235/9N210 Cluster Submunition Detector,"
The Journal of Conventional Weapons Destruction: Vol. 27
, Article 9.
Available at: https://commons.lib.jmu.edu/cisr-journal/vol27/iss2/9