Drone Can Identify Violent Behavior Through Deep Learning Tech

This post is also available in: עברית (Hebrew)

Police and other governmental agencies around the world have been employing drones to be their eyes, searching the horizon for crime or terrorist activity. Drones can be found in war zones monitoring hostiles or in the air spying on foreign drug cartels. Cambridge researchers have taken this notion one step further and figured out a method to classify live footage from the drone and notify authorities of suspicious and violent behavior.

The project was undertaken by several researchers and is entitled ‘Eye in the sky: real-time drone surveillance system (DSS) for violent individuals identification’ using scatternet hybrid deep learning network (a hybrid deep network between Bluetooth-enabled devices).

The drone uses pyramid networks to identify the humans and focus on them. It uses scatternet to help the drones analyze the footage and let the authorities know what the situation is.

In the eye of the drone, the human form is broken down into 14 points from head to toe. These points are connected by lines that signify the arms, legs and body. The drone can tell, based on quick measurement of the angles, if the subject is potentially dangerous.  

Analyzing these videos proves difficult due to illumination changes, shadows, poor resolution, and blurring, but the recent tests have proved quite resilient, according to designboom.com.