Research On Lane Detection Technology Based On Vision

Currently, research on intelligent vehicle technologies is becoming hot in many fields, and has attracted much interest from the research scientists. Of these technologies, vision based vehicle navigation is one of the most significant areas, and it mainly concentrates on the problem of how to detect road lane information. The work of this thesis is given as follows.Firstly, the background and significance of the road lane detection technology using computer vision is introduced, as well as the advances of intelligent vehicle research at home and abroad, and two conventional road lane detection methods, namely model based method and feature based method, are studied in detail.Secondly, image pre-processing algorithms are presented for the vehicle camera video, including color image gray conversion, histogram equalization, and median filter to realize image noise filtering. A self-defined difference operator is then proposed to extract the image edge, by the sum of a pixel and its eight adjacent pixels with different weights. Experimental results show that the self-defined difference operator is able to achieve better performance to obtain image edge features. Meanwhile, region growing algorithm is presented here to extract the regional features of an image by region division.Then, an algorithm integrated edge and region features is proposed to implement edge segmentation, using the complementarity of their features, and then the line parameters are obtained using Hough transformation. Using the slope of road lanes and the intersection angle of two lanes in an image, the lanes can thus be recognized.Finally, plenty of experiments were carried out under varied illumination and complicated environment conditions. These results show that the algorithms proposed in this thesis can be used to identify the road lanes from the video fastly and robustly.