Real-time lane recognition by simulated annealing algorithm

In this paper, an algorithm to recognize driving lane from road image is presented, which is used for an autonomous driving vehicle. The algorithm is applicable to lateral control of the vehicle in real-time. The first stage of the process includes grey level digitization, the edge detection of white lane markers and inverse perspective remapping. This is followed by lane detection process which decides parameters of quadratic polynomial assumed as a road model. These parameters of the road model are determined by maximum likelihood parameter estimation method, which has advantages in estimating variables of maximum probability. Also, the estimation method is implemented by metropolis algorithm on computer system. The information of driving lane can be obtained to the driving system every 0.035 seconds. Also, it has been shown to robustly recognize driving lane even in the presence of shadow, discontinuity of lane marker and varied lighting condition.