Research On Lane Detection Technology For Expressway

Vision-based navigation is one of the main technologies in today's driver assistance and unmanned driving system, with its advantages of wide range of signal detection, complete environmental information, lower costs and consistent with the human perception. In-depth research work was done on the key technology in visual navigation system-lane detection with the purpose to develop a lane detection method applied to navigation system for automated vehicle on expressway. This thesis mainly includes the following:Some image preprocessing methods were discussed to find a preprocessing method which can resist rain, fog and other weather conditions,and have the ability to eliminate redundant information, and highlight lanes.Summarized and analyzed existing lane models and related hypothesis.After discussing how to locate vanishing point and line, a rapid offline vanishing point location method base on camera parameters was proposed.Vanishing point was updated according to lane detection results,and used to protract lanes which were partially hid.A lane detection method based on road region segmentation and Probabilistic Hough Transform was proposed.This method firstly segments the region of road, and then detects lanes in road area. Among them, road area segmentation using color features extraction in triangle area and connected region statistics is also suitable for suburban road. Applying large quantized interval and lane length threshold to improve Probabilistic Hough Transform can effectively solve the undetected problem caused by lack of lane edge pixels.Dashed lane is critical information for expressway. To satisfy the requirements of road partition, apply simple statistical theory to line style judgment and then get the driving region rapidly and accurately.The results of experiments show that the proposed method can effectively solve some issues in lane detection, and it is real-time and robust.