Abstract: A new method for lane detection in uncertain circumstances is proposed. After determining the color scale of the road, the color information of the lane marks is extracted to eliminate the influence of the illumination changes. The interference of the vehicles with the lane detection is eliminated through comparing the size, shape and motion pattern between the vehicles and the lane marks. By storing the single-image lane detection results in the lane-coherence-validation module, the accuracy of the lane detection enhances. Experimental results show that the proposed algorithm is able to robustly find the lane boundary under various lighting conditions and to solve the problem of the occupancy of the road by the vehicles.