A lane detection method based on track management approach

The lane detection is important for the lane departure warning (LDW) used in advanced driver assistance systems (ADAS). Several approaches for lane detection were suggested in the past. However, there is still one issue about robustness. This paper presents a robust lane detection method based on a two layer Kalman filter. The key idea is to apply methods from the target tracking domain to identify lanes in the image space. We assume each road lane could be seen as one track in 2D coordinates and then use a Kalman filter to update the tracks. Using information from consecutive frames, it is much easy to distinguish between real lane pixels and noise in the image. Our method is based on two phases: an image preprocessing phase to extract areas that potentially represent markings and a tracking phase to identify lanes. In the tracking phase we use two special Kalman filters to estimate tracks according to the track management in spatial dimension. The simulation results show that this method exhibits good robustness under various scenarios and meet the real-time requirement.