Real-time vision for driving supervision

摘要

The aim of this work is to propose a computer vision system working at video rate for driving supervision. More specifically we focus on the problem of vehicles running off the road. The system has been designed to make a real time analysis of road images, to determine the position of the white and yellow lines used to mark the road limits and the position of the vehicle relative to these lines. The architecture is composed of 3 main modules: the first module makes a real time color segmentation of the yellow lines. The second hardware module is a specific DSP for straight line detection in any direction. The final module includes a specific algorithm that decides if the line markings at the side of the road are in the zone for safe driving.