Thanks to AI-powered machine learning, automatic observation is becoming more and more efficient. A project by British and Indian scientists has demonstrated the new capabilities of this technology - in particular, it allows you to identify aggressive people in a crowd using drones.
In an article titled “Eye in Sky, ” the researchers describe how they used a conventional Parrot AR quadcopter costing about $ 200 to stream video over the mobile Internet for real-time analysis. The algorithm, obtained as a result of intensive machine learning, made it possible to identify aggressive-minded individuals by the characteristic elements of their poses. Five positions related to violence were chosen for training - strangulation, punches, kicks, shooting and the use of piercing objects.
Scientists hope that such algorithms will find their application in solving crimes in public places and during public events. Amarjot Singh, a leading researcher at the University of Cambridge, cited the 2017 Manchester Arena bombing as an example. According to him, such attacks can be prevented if CCTV cameras record suspicious behavior - for example, leaving a bag unattended for a long time.