Abstract :
Lane detection from a live video captured in a moving vehicle is an important issue for autonomous vehicles and video-based navigation systems. In this paper, we present a novel idea for robust lane detection and lane color recognition. More specifically, a framework for robust lane detection is presented. Then, a novel idea to reduce illumination effects is presented. Lastly, SVM approach is presented to recognize lane color robustly for various lighting conditions including shadow, backlight, sunset, and so on. By combining information from navigation database, it is possible to decide if we are in the leftmost, middle, or the rightmost lane, which allows us to provide more realistic navigation information to drivers. Simulation results are provided to show the robustness of the proposed idea.
Keywords :
image colour analysis; navigation; object recognition; support vector machines; video signal processing; SVM; lane color recognition; lane detection; live video; moving vehicle; navigation database; realistic navigation information; utonomous vehicles; video-based navigation systems; Data mining; Intelligent vehicles; Lighting; Mobile robots; Navigation; Remotely operated vehicles; Roads; Robustness; Support vector machines; Vehicle detection;
