| This article deals first with
a process designed to detect the circulation lane of a vehicle by onboard
monocular vision. This detection process is based on a recursive updating
of a statistical model of the lane obtained by a training phase. Once the
lane has been located, a reconstruction algorithm computes the vehicle location
on its lane and the three-dimensional shape of the road. Thereafter, the
authors seek to detect, locate, and track front vehicles equipped with specific
visual markers in order to achieve an accurate determination of their location
and speed. By combining these various data, the most dangerous obstacle
can be identified. Each of these three processes is described in detail,
and significant examples are provided. |
