Lane detection method based on moving blurring feature and WMSE

A fast and robust lane detection method was presented. Firstly, we limited the detection operation in the region of interest according to perspective imaging theory, then 4 kinds of edge points are extracted based on the character of gray change of strong lane edge after moving blur. The lanes are width enough, so each lane can obtain 2 separate edge point sets. They are close to each other, almost equal in area, and consistent in direction. We can remove non-lane edge point by these features. Finally, we got the parameters of lanes by means of line fitting of weight minimum square error (WMSE) method. The experimental results show that the average detection rate is about 93.6%, and the average processing time is 35ms per frame.