With satellites falling out of the sky, the US military is beefing up its surveillance capabilities using robotic unmanned aerial vehicles (UAVs).

The latest development is a new kind of optical sensor using hyperspectral imagery, being developed by scientists at the Rochester Institute of Technology in the USA.

The sensor automatically choose three different modes - black and white imaging, hyperspectral or polarization - depending on its target.

Monochrome magic

If it's tracking a car, for example, the sensor will collect a black and white image and record its shape. Then a hyperspectral image will plot the object's colour in multiple wavelengths, from the visible light to the near and short infrared.

The third imagery mode, polarization, cuts through glare and allows it to distinguish between objects of similar colour and shape.

"The idea is that if the object you are tracking goes into an area where you lose one piece of information, the other information might help," says John Kerekes, an Associate Professor at the RIT.

Kerekes and his team are testing their preliminary models using generic scenarios played out in a simulated world akin to Second Life.

1800MP digital camera

In other UAV news, Wired reports on a 1.8 gigapixel eye in the sky being built for the US military. The ARGUS-IS system will be able to open 65 separate video 'windows' over an area of 100 square miles, tracking individual cars or even people.

ARGUS-IS (standing for Autonomous Real-time Ground Ubiquitous Surveillance - Imaging System) will generated around 27GB of data every second.

It has been designed to be mounted on a new unmanned helicopter called the A-160 Hummingbird that can hover for up to a day at an altitude of over 15,000 feet.