There are several ways to offer night vision to a drone, but a group of researchers from the University of Zurich and the NCCR Robotics is creating a rather original one.
It is an eye-inspired camera that can work when the drone moves at high speed, and even see in near-dark conditions.
The goal is not to rely solely on the GPS, nor to have a lot of light in your environment, nor to have to go very slowly in order to obtain the necessary recognition of the environment, nor to depend on extremely expensive laser scanners. The new system is made up of sensors inspired by the human eye, which detect changes in brightness at the pixel level instead of frames of standard intensity. The camera retina does not require full light capture to generate a clear image.
Professor Davide Scaramuzza, Director of the UZH Robotics and Perception Group, comments that it is the perfect solution so that drones can be truly autonomous regardless of light conditions. On how the system works, they indicate:
Traditional video can be divided into a series of frames containing rich pixel-level information on brightness and color. Event cameras, by contrast, only compare the brightness in each pixel from one moment to the next.
Such a camera does not move, so useful information is very scarce, but its readings can be analyzed by computer to visualize an environment. The UZH researchers designed software to efficiently process the output from these cameras, and the price would be much cheaper than that of laser sensors.
There is still a lot of work to have this type of camera in commercial drones, but it is a good first step.
