Research On Lane Detection And Recognition Algorithm Under Complex Road Conditions

Lane deviation caused by the driver’s inattention or visual disturbance is one of the important reasons of traffic accidents. Advanced driver-assistance systems(ADAS) can reduce the occurrence of traffic accidents, and has very important practical significance. Lane detection is not only one of the core technology of the system, but also an important prerequisite for the realization of vehicle safety driving assistant. In recent years, domestic and foreign scholars have made some significant lane detection algorithms, but two problems remained. The first problem is how to improve the algorithm’s robustness in a strong interference environment.The second is how to improve the real-time and not affect the robustness in the environment.In order to improve the real-time and robustness of lane recognition algorithm,a feature extraction algorithm based on bilinear interpolation is proposed. The principle of which is based on the fact that only the points belonging to the lane markings will keep on the same straight lines by zooming around the vanishing point of the lanes. A new gradient-enhancing conversion method is proposed for the low contrast between road and lane. Linear discriminant analysis was been used to calculate the optimum conversion coefficients of the current image, so that the gradient value between the road and lane is large sufficiently. In addition, the optimum thresholds can be calculated by using the adaptive Canny edge detector which could exclude the outliers and fit the lane effectively. Lastly, be used for the fitting lane.An improved Hough transform algorithm is used for straight line fitting. The partition fitting lane detection algorithm can be used for straight line or curve line.Many experiments are done and analyzed under various road conditions, the results shown that: the feature points can be extract efficiently by using the bilinear interpolation algorithm when the lane is discontinuous, shaded or uneven brightened. At the same time, the algorithm based on gradient-enhancing conversation has good performance when the contrast between lane mark and road is low.Using the improved Hough transform algorithm for laneline fitting can extract the lane line quickly and accurately. The partition fitting lane detection algorithm can identify the lane mark quickly and accurately in a variety of conditions. The detection accuracy of the algorithm is as high as 88.5% when the extraction effect of feature points is better.