Design And Implementation Of An Algorithm For Lane Tracking And Detection Based On Hyperbola Model

With the development and popularization of computer technology, more performance,more professional and smaller of the computer system is applied more and more widely. Inthe automobile manufacturing field, it makes the small volume and strong security intelligentvehicle auxiliary driving system possible. Auxiliary driving system evaluates the drivingcondition through analyzing the state of driver behavior. Before the danger happens, systemmust alert the driver. Lane detection is a necessary security link in intelligent vehicle anddrive assistance system. Accurate lane detection and mathematical fitting can remind whetherthe vehicles has driven across the line while ensure the lane detection-based security systemequipped with the early warning capability.This paper summarizes and compare various lane detection algorithm, and then putforward a lane detection method that based on Hough transform and piecewise hyperbolicmodel of Kalman filter. Compared with the traditional methods, our method has a good effectin curve lane detection. Our algorithm detects the edge with the Sobel operator in special area,and extracts the lane straight edge. And then detect the straight line of lane edge in near fieldwith Hough transform. According to the results, we use weighted hyperbolic model andmatching the shortest distance method to fit the lane in far field. Through segment of lanedetection and accurate judgment of curve direction, we fit the curve line whose curvaturechange biggest through changing the hyperbola parameters. After that, we use Kalmantracking to constraint the lane detection range and improve the detective rate. Moreover, wedesign the lane curvature filtering mechanism which can get the fittest curvature parametersof hyperbolic model and eliminate the false detection results effectively. In the experiment,we find that the main reason interfere the effects of lane detection is the vehicles on the road.To improve the effect of detection algorithm, this paper adds the detection of obstacledetection, in order to remove the interference before lane detection. Our algorithm canincrease the detection rate.The experiment confirmed that our algorithm has good detection effect and processingspeed and it can meet the real time detection of all-weather running of straight and curve lane. Our lane detection algorithm is not depending on the light and the color information of lane.At the same time, the Hough transformation method also has a good effect in discontinuous orpartially incomplete lane detection. It can divide the instate area effectively and excluded theinterference from non-road and vehicle. It has robust performance and has good prospect ofindustrial application.