Real-time lane detection-based line segment detection

Abstract

This paper introduces a robust algorithm for real-time lane detection using the lane markers in urban streets or highway roads. It is based on applying Region of Interest (ROI) on the input image of the road from a calibrated camera in the front of the car, generating the top view of the image using Inverse Perspective Mapping (IPM), applying the core algorithm Line Segment Detection (LSD) which is followed by post-processing steps. Applying curve fitting to the line segments to get the right and left lines or curves. Finally, to get the output stream inverse IPM is applied. The proposed algorithm can detect the road lanes discriminating dashed and solid road lanes, straight and curved road lanes overcoming the shadow effect challenge with real-time performance 70 frames per second.