Research On The Lane Detection Algorithm For Express

Lane detection is that the continuous or discontinuous lanes in the driving regional arereconstructed two continuous straight lines from the images which contains lane marks. Lanedetection is the premise and basis of the whole Lane Departure Warning System. Whether thelanes in the road image can be identified quickly and accurately or not, it is important for thestability and the accuracy of the warning system. Although, there were many valuableresearch results proposed in recent years. But because of the complexity of road environmentand the influence of many kinds of bad weather, the road images captured by the vehiclecamera are different. So, how to improve the performance of real-time and robust wasbecame the focus of the study related. There are still significant issues needs to further study.In order to extract lane line of different environmental conditions, firstly, the grayimages are processed by median filtering, and the interest region is extracted. Then accordingto the characters of roads, the images are processed by different corresponding algorithms.The road images have high definition under normal illumination, in order to get the featurearea of lane line from road image, the images can be directly segmented, the small area noisesare removed and the feature points are extracted. After processed by global Hough transform,the lane line parameters can be obtained. Under Weak illumination, the contrast between laneand road is not high, the image can be enhanced from taking the±45°Sobel operator; if thepavement is covered by the shadow, the lane line need to use the improved±45°Sobeloperators to enhance; if the road was covered with rain, the lane line feature is not apparent,the lane line can be enhanced by the steerable filters; if some of the road details areobscured by the fog, firstly the road images need to be equalizated, and then the lanes areenhanced by the steerable filters. After pretreatment, the road image needs to be segmentatedby the Two-Dimensional Improved Otsu algorithm. Finally the lanes’ parameters areextracted by Zoning Hough Transform, and the lanes are fitted out.The experimental results under various conditions show that the road images underdifferent weather are processed by different algorithms, so the adaptive can be improved. A lot of noise can be removed by using area threshold, the real-time and robustness of algorithmcan be improved. After being enhanced by the improved Sobel operators, the lanes techniquesintuitively are presented linear features, and there are more feature points of the lanes, so thelanes’ parameters are extracted more easily. Compared with the global threshold quantity andthe local threshold quantity, on the premise of ensuring real-time, the anti-noise performanceis improved. Using Zoning Hough Transform, the interference of the ambient noises can beeliminated, and the subjective position can be prominented. The parameters are obtained bythe comparison of the same lane in different regionals and then the lanes are identified. Theaccuracy rate of Zoing Hough Transform is bigger than the one of Global Hough Transform.This lane detection method can highlight the characteristics of lane, extract the lane markparameter accurately, and increase the accuracy of the identifying lane.