AI, target tracker, microcomputer: Ukrainian FPV drones with computer vision

Ukraine is preparing for production of a new domestic solution with machine vision for FPV drones, which is already at a high stage of readiness and is being actively tested by the Dyki Shershni (Wild Hornets) community, Defense Express reports.

According to the developers, this is a software and hardware solution based on a Ukrainian-made microcomputer (of course, assembled from off-the-shelf components). But the software is the most important thing in this case, as it is a neural network and a target tracker.

This technology allows processing video images, recognizing the outlines of objects — people, vehicles, armored vehicles, artillery, etc. — as well as capture and attack them without human intervention.

This is the main advantage of machine vision drones, as they eliminate dependence on electronic warfare, as the last 100 meters of the FPV drone is already flying independently. In addition, this technology overcomes the dependence of attack effectiveness on operator skill.

Of course, autotracking technology is not a know-how, but the main challenge is to make it work on the cheapest microcomputer, which should also be small. The software itself, which has to work flawlessly, also has quite high requirements, and given that it is a neural network, it really needs to be trained to recognize shapes in different conditions.

Photo: Wild Hornets

"In the near future, we will have a lot of work to increase the combat potential of the Ukrainian Armed Forces. But for this, we will continue to need your support," the Wild Hornets community said in a statement.