New night-vision system to bring down road fatalities
Researchers at the University of Granada (UGR), Spain, have developed a microchip that improves night vision for drivers. The system complements the illumination of the car and will alert
drivers to obstacles on the road, the scientists explain. The device was invented in the framework of the EU-funded DRIVSCO (Learning to emulate perception action cycles in a driving school
scenario) project, in which the UGR participates.
The new system will extend the field of vision beyond the range normally illuminated by conventional headlights by using information extracted automatically from night visors. Information about
movement or depth is provided in real time by two infrared cameras installed on the car, so that, for instance, bends in the road, pedestrians and other cars will be more easily detected. The
new microchip facilitates the extraction of information from the cameras. In addition, it can be used to set off an alarm – by visual, acoustic or other means – to alert the driver.
The system could, in the future, help to reduce road deaths: Four out of ten fatal car accidents occur at night, although there is about 60% less traffic then than during the day. This is due
to reduced visual acuity and field of vision, as illumination is usually insufficient for ideal vision. ‘Dipped headlights only illuminate about 56 metres when the breaking distance at 100 km/h
is about 80 metres,’ says Eduardo Ros Vidal from the UGR, the researcher who carried out the study.
DRIVSCO is a project funded under the Sixth Framework Programme. It investigates real-time vision and its application to the car industry. Researchers are focusing, first and foremost, on
night-vision scenarios with infrared illumination as the commercially most relevant application domain, they say. DRIVSCO brings together universities and industry from Spain, Lithuania,
Germany, Italy, Denmark and Belgium in a joint effort to design intelligent cars that could make driving easier and safer.
The central idea is that cars should learn to drive autonomously by correlating information about the environment and the actions of the driver. Starting on the basis of a fully operational
human-machine interface, such a system should be largely independent after learning, using a variety of predictive mechanisms. ‘We envision a system that can learn to drive a car during
daylight and apply the learned control strategies in an autonomous way to the system’s augmented field of infrared night-vision,’ the DRIVSCO researchers say.
For further information, please visit: