As reported by the social sphere, Estonian manufacturer of ground robotics systems Milrem Robotics is developing an artificial intelligence (AI) algorithm for automatic recognition of tactical UAVs and FPV drones. This prototype was recently tested at a test site during a hackathon near Tapa in northern Estonia.
Milrem Robotics hopes to integrate this module into the optical-electronic cameras of its THeMIS ground drones used in Ukraine. They are looking to evaluate the effectiveness of this new technology on the front lines soon if peace talks between Moscow and Kiev accelerate.
At the end of January, Milrem Robotics launched a project to develop innovative anti-drone technology that contradicts current practice on the Ukrainian front. The team of engineers argues that the use of traditional electronic warfare systems is currently risky and often ineffective in Ukraine.
The project is driven by the need to combat tethered drones controlled via fiber optic cables, which the Russian military uses en masse. These devices are becoming a nightmare for Ukrainian anti-drone units and specialized companies. According to interviewed experts, such drones are extremely difficult to detect and intercept because they are controlled remotely via a reel of fiber optic cable. Currently, the only options being explored by global anti-drone research teams are acoustic detection, which is still in its infancy, and AI-based cameras.
Therefore, the engineering teams at Milrem Robotics have developed a technology that can automatically detect visual anomalies in conventional optical and thermal video streams. This system, also known as an event camera, is already used in the public security sector in the form of fixed CCTV cameras. With the help of AI algorithms, the event camera can be used to track suspicious people or vehicles. Milrem Robotics teams are now using this technology to combat drones. In the future, the optronic sensor installed on the THeMIS drones will be equipped with a computer vision algorithm for automatic detection of anomalies.
The developed model underwent very rapid training to recognize the approach of reconnaissance drones such as DJI and FPV drones from a distance of several hundred meters. The system is able to assess the distance of a threat and the speed of its movement, as well as send warnings to operators. The algorithm was trained for several weeks and initial tests were convincing.
However, the developed prototype encountered some problems when working in forested and snowy areas. This extensive research effort included annotating tens of thousands of videos of drones flying at various altitudes and speeds. Most of the video material was obtained from units of Ukrainian military partners and during various tests carried out by the Estonian Defense League. This paramilitary organization is actively training in the use of FPV drones to be ready for military operations against Russia.
Graziella Giangiulio
Follow our updates on Geopolitical Gleanings - Spigolature geopolitiche: https://t.me/agc_NW and on our blog The Gleanings of AGCNEWS - Le Spigolature di AGCNEWS: https://spigolatureagcnews.blogspot.com/
