Abstract: Aiming at the problem that it's difficult for traditional Canny edge detection operator to extract the lane line features robust in real-time in complex scenes of a large amount of traffic information, an improved method for the optimal direction interval quick detection based on Gabor filter was proposed. First, the panoramic image was expanded into a rectangular image by the concentric circle approximation expansion method. Then the expanded image was processed by Gabor filter with different phases, thus obtaining the direction interval which had the highest clarity of the lane line quickly. In the process of edge detection with Canny operator, only the edge points in the optimal interval were used for non-maximum suppression and further processing. Finally, the algorithm was tested on 500 real images for videos, and the recognition rate could be better than 94.2%. The results show that the proposed method is robust for complex environment and has strong feasibility, which can effectively improve the real-time performance and stability of lane departure warning system.