Daimler announces a new driver assistance system for the city vehicles which will be capable of recognizing and viewing dangers the way a human does. The software will label scenes that are seen on the road and analyze for danger signals. For instance, identifying the different behaviors of the pedestrians on the road and anticipating their movement. Lane changing will also be easier with this system. Other capabilities of the system include understanding where to change lanes or when to turn.
City traffic offers diverse situations on the road. There are cyclists as well as pedestrians, cross traffic and pedestrians who are engrossed on their smart phones or children accompanying parents. These situations have varying levels of risk inherent in them which need to be identified instinctively by a driver when driving along a busy city road. These situations can be predicted and made easier to handle by a driver assistance system which Daimler is planning on launching. The company claims that driving on urban streets will be less stressful and safer with the help of this kind of a system assisting the drivers.
The system is called labeling of scenes which is a camera based support system being developed for the drivers. This system will classify the different situations and identify the different objects that need attention from the driver. These could be pedestrians crossing the street or wheelchair bound people or cyclists.
The researchers who are working on the system called environment sensing have identified about twenty five different object categories which are common situations found during urban driving. The system learns based on these inputs on how to classify the different scenes witnessed on the road and it can detect objects that need to be seen by the driver which might be far away or hidden. The deep neural networks are the artificial neural network system that is being used to develop this assistance program.
The system will work complimentary to human sight. It will work as the human sight does, taking in the sights and using the neural processing to identify the objects, the situations and risks that might be posed by them. Scene labeling allows the inputs from the camera to be interpreted as the eye and the brain does in human beings. This is made possible by the advances that have come about in computing power which can mimic the neural functions of the brain and be able to interpret different scenes as the human mind would. Hence, even complex scenes in an urban traffic can be interpreted and risks or dangers highlighted for the assistance of the driver by this system.
Image Credit : FreeScale