Research On The Real Time Lane Detection Based On Monocular Vision

Driver safety has been an area of interest with the sharp increase of the number of the vehicles. Automation of vehicle driving is being developed rapidly nowdays due to the consumer demand. An important component for intelligent driver assistance system is the lane detection module. This paper collects road scenes by camera, and then extreact lane edge information through a series of visual analysis and therefore warn a driver of an approaching danger.This thesis does research on the real time detection of lane markers.The main works have been accomplished are as follows:1. In order to solve the problem of the false lane detection because of the the road markings, the lane pre-processing method is based on temporal blurring and IPM(Inverse Perspective Mapping).The captured color images undego temporal blurring by averaging three successive frames.This smoothing helps connect dashed lane markers to form a near continuous line. IPM transformation is to take generate a top view of the road image and can get rid of the perspective effect.Under the condition of the road markings, it can realize the correct lane detection.2. Based on the steerable filter, the transformed IPM image is filtered by a two dimensional Gaussian kernel for the vertical direction and the horizontal direction respectively. It keeps the actual pixel values of the thresholded image. This approach uses the lane orientation information and can detect more lane pixels. It is useful to the post-processing lane fitting.3. Due to the Hough transform has detected more lanes, the model single and the RANSAC(Random Sample Consensus) has the large amount of computation, In the first place, the paper uses the Hough Transform in order to detect lines where the lane may exist and it is rough detection.Then, it focuses on a window around lines and run the RANSAC fitting on that window to find the lane true position.4. Due to the road breakage and dashed lane markers, it cannot detect the lanes in some frames. This paper puts forward a scheme of lane tracking algorithm.The parameters of left and right lane are predicted using Kalman filter. According to the different situation of the lane(dashed, solid or temporary disappear), apply different scheme to resolve the problem.The algorithm can effectively detect the lane marker and resolve the problem of the leak lane detection.