A Novel Lane Detection Algorithm Based on Support Vector Machine

Abstract

In this paper, a new lane detection algorithm based on support vector machine (SVM) is presented. This algorithm can overcome the flaws when applying traditional lane algorithms which are only applicable to some special situations. The main steps of this algorithm: road surface extraction by using SVM pattern recognition, image morphology operation, transforming the image into a bird-view image by using the relationship of image coordinate system (ICS) and world coordinate system (WCS), getting the center points from the road’s midline, regressing the road shape function by using SVM. Relative to the traditional laser-based sensors and millimeterwave radar which are classified as active sensors, vision-based passive sensor has an advantage that it acquires data in a noninvasive mode and will not alter the environment.