Detección y modelado de carriles de vías interurbanas mediante análisis de imágenes para un sistema de ayuda a la conducción

摘要

The goal of this work is the study and design of a Lane DepartureWarning System. The system has been implemented on the research platform IvvI ( Intelligent Vehicle based on Visual Information ). The system analyses road image sequences captured from an on-board video system. The aim is to anticipate dangerous manoeuvres and warn the driver, for instance, when the vehicle is about to cross a solid line, or before an unintended lane crossing. This is achieved by means of the combined work of four algorithms. A rst algorithm detects and tracks the road lines. A second algorithm classi es these lines into three di erent classes (solid lines, dashed lines, and merge lines). A third algorithm detects the lane change manoeuvres and warns the driver if the manoeuvre is dangerous or unintended. Finally, a self-calibration algorithm is able to calibrate the vision system without the need of arti cial partners. The whole processing is performed in real time. The system has been tested under real conditions on the IvvI platform. __________________________________________________