Computer vision-based multiple-lane detection on straight road and in a curve

Abstract

A vision system was designed to detect multiple lanes on structured highway using an “estimate and detect” scheme. It detected the lane in which the vehicle was driving (the central lane) and estimated the possible position of two adjacent lanes. Then the detection was made based on these estimations. The vehicle was first recognized if it was driving on a straight road or in a curve using its GPS position and the OpenStreetMap digital map. The two cases were processed differently. For straight road, the central lane was detected in the original image using Hough transformation and a simplified perspective transformation was designed to make estimations. In the case of curve path, a complete perspective transformation was performed and the central lane was detected by scanning at each row in the top view image. The system was able to detected lane marks that were not distinct or even obstructed by other vehicles.