Font Size: a A A

Research On Vision-based Lane Markings Detection Technology

Posted on:2018-07-07Degree:MasterType:Thesis
Country:ChinaCandidate:C Z HouFull Text:PDF
GTID:2322330515971174Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the rapid increase of car ownership,the urban traffic is becoming more and more crowded,which leads to people's widespread attention to the intelligent transportation system(ITS).Lane markings detection is one of the critical technologies in intelligent transportation field,which is widely used in driver-assistance system,lane departure warning system and vehicle anti-collision system,and it is significant for improving the traffic safety.Vision-based lane markings detection may be easily affected by some interference from external circumstances such as the illumination change,shadows of trees,writings on the road and the passing vehicles.To solve these problems,we use line segment as low-level feature to analyze the structural information of lane markings,and a lane markings detection method under vanishing point constraints is proposed in this thesis.Firstly,the region of interest(ROI)is divided dynamically according to the mean value of road image in each row,and the edge information of road image is extracted from the ROI.Then,we filter out noise edges with abnormal orientation based on the orientation-priority searching method.Finally,candidate line segments are extracted by progressive probabilistic Hough Transform(PPHT),and the no-lane markings are eliminated under vanishing point constrains to realize the lane markings detection.In order to improve the adaptability of road model to lane markings with different shapes,the Bezier spline model is applied to construct a deformable template by random sample consensus algorithm.Firstly,the top view of the road image is obtained through inverse perspective mapping,then filtering is performed on the top view space in vertical and horizontal orientation respectively,and this operation can reduce the interference of noise and enhance the information of lane markings.During the fitting step,a deformable template is constructed with third-degree Bezier spline model by random sample consensus algorithm,and the parameters of spline model are adjusted according to the actual shape of the lane markings,which is able to fit lane markings with different shapes.Through the test of road images under different scenes,the experimental results demonstrate that our method can not only overcome the interference from external circumstances under complex road scene,but also detect lane markings with different shapes,and it shows a good accuracy and robustness.
Keywords/Search Tags:Vanishing point constraint, Noise edges, Progressive probabilistic Hough transform, Random sample consensus algorithm
PDF Full Text Request
Related items