Study Of The Vision Based Lane Detection System

This paper presents a real time vision-based lane detection method to detect lanes by using an on-board monocular camera fixed at the front of the vehicle.In the proposed method, two levels of processing are involved. One is a single frame level,one is an inter-frame level.At the single frame level, after down-sampling the video image, a grid-based morphology operation is firstly used to generate regions of interest(ROI), in which lane-marks are searched for by scanning for an edge-link pair group that corresponds to one lane-mark. During this process, an extended edge-linking algorithm with directional edge-gap closing is used to produce more complete edge-links,and features like lane-mark edge orientation and lane-mark width are used to select candidate lane-mark edge-link pair groups.For final verification of the actual lane-mark regions, color is checked inside candidate edge-link pair group regions using the YUV color space, on which detected lane-marks can be classified into yellow, blue and white lanes.Additionally, the continuity of the lane is estimated employing a Bayesian probability model based on lane color and edge-link length ratio.At the inter-frame level, lane parameters acquired at the single frame level are stored and analyzed so as to estimate the position and type of lane more reliably.