The Real-time Lane Detection By Parabola Fitting Based On RANSAC And Type Identification Algorithm

With the development of society, the accelerated pace of life, the vehicle is an important tool for people to improve efficiency and convenient life, but also brings more security risks and traffic problems. The intelligent vehicle driver assistance systems can alert drivers and help to make the right operation, reducing traffic congestion, accidents and other phenomena, people’s life will be more intelligent by the driverless vehicle.The lane detection is part of the lane departure warning system, through the on-board camera incoming video frames after a variety of image processing algorithm, it can detect the line appear in the image on the real-time, when the vehicle go through the line, it could remind drivers in time to provide protection. The lane detection also play an important role in realizing unmanned vehicle, which included in the auxiliary driving system. The driver-less car mainly through a variety of radar scans the obstacles, vehicles, roads and other information, using path planning algorithms, determine the car driving route, and the automobile motion state after processing, such as speeding up, slowing down, stop, turn, turn around, such as automation, by on-board camera and computer vision technology to detect lines, vehicles and other information, using images to achieve unmanned lonely cost is very low, but cannot reach the actual operation of the accuracy requirement. Because of the accuracy of radar measurement, it’s priority to calculate the road information by radar, the image processing technology is complementary, to realize vehicle driver-less.In this paper, using the on-board camera to get the images as input, after smoothing filter firstly, then according to the position and internal reference of camera to Inverse Perspective Mapping, using Canny operate the projection results for edge detection, then to detect the straight line in the edge image by Hough transform. Lines are separated into groups according to their middle points positions for producing the external rectangle of every group, using the random sample consensus to parabolic fitting processing of the key points in the rectangle area. Finally, judge the lines type after white balance operator by the equal step color comparison, the proposed algorithm can detect multiple lanes lines in one frame, and identifying the lane types, including(white solid line, white dotted line, yellow solid line and yellow dotted line), and it also can meet the real-time requirements.