An Efficient Lane Markings Detection and Tracking Method based on Vanishing Point Constraints

This paper proposes an efficient lane markings detection and tracking method, which utilizes line segments as feature information combined with vanishing point constraints. The method can be highlighted in four items as follows. Firstly, the region of interest (ROI) of road image is determined, and then the edge information is extracted by Canny edge detector. Secondly, edge image is scanned to calculate the orientation of edge-linking pixels and we filter out noise edges which have abnormal orientation. Then, line segments detected by Progressive Probabilistic Hough Transform (PPHT) are applied to analyze the structural information of lanes and the interferential line segments are eliminated under vanishing point constrains. Finally, K-means clustering algorithm is used to classify and fit the closest two lane markings. Specifically, a Kalman filter is utilized for lane markings tracking. The experimental results demonstrate the good accuracy and robustness of our method in various complex environment.