Vision-based lane recognition under adverse weather conditions using optical flow

Lane recognition is the basis for many driver assistance systems, including lane departure warning (LDW), the assignment of vehicles to specific lanes and fully autonomous driving. A major problem of common vision-based lane recognition systems is their susceptibility to weather. Especially when driving in adverse weather conditions such as rain or snow it is difficult to estimate the road course. The contrast between the white lane markings and the pavement is poor, sometimes the colors of the markings are negated. Furthermore the range of sight is reduced enormously causing a bad prediction of the lane parameters, particularly the curvature. We present a solution which relies not only on finding white markings. Correlating structures parallel to the road over time the horizontal optical flow is calculated. It is then integrated in the lane recognition system estimating the position of the vehicle within the lane and the curvature parameters of the road ahead. The system allows to assign obstacles detected by radar or vision reliably to specific lanes and to drive laterally controlled autonomously even under adverse weather conditions.