Collision-avoidance system based on optical flow

Proposes a collision-avoidance system for vehicles, which warns the driver of impending collision. It is based on the parallel computation of the image-motion velocity vector field over a large part of the image plane. This image-motion field is then used as the input to a neural network which computes two properties of the movement of the vehicle relative to its environment. The first is the direction of heading of the vehicle in terms of azimuth and elevation angles. The second is the time-to-contact with the immediate environment in front of the vehicle given the current velocity of the vehicle. The detection of these two properties could be used to warn the driver of impending collision or could be used to make online adjustments to the current trajectory and path of the vehicle