Drones could help detect crime by spotting violent behaviour in live footage.
Scientists are currently researching a system which could allow the camera equipped devices to asses a stream of a crowd, and pick out people behaving in an aggressive manner.
In a paper entitled 'Eye in the Sky', researches outline their plans for a Parrot AR quadcopter - costing around $200 - to use a mobile internet connection to transmit videos for real-time analysis.
It would use an algorithm using deep learning to estimate poses of humans in the footage, and match them to postures designated by scientists as "violent", as reported by The Verge.
Five poses have been included for the purpose of the research project, which are strangling, shooting, punching, kicking and stabbing.
According to lead searcher Amarjot Singh - from the University of Cambridge - the project has been motivated by terror attacks such as the Manchester Arena bombing last year.
However, the accuracy of the technology remains to be seen, as it currently decreases when more people appear in the frame.
There is also the issue that the research doesn't represent the software's use in the real world, as it has been tested in controlled conditions with volunteers pretending to attack each other.