Real-Time Driver Assistance Systems via Dual Camera Stereo Vision

Abstract

This paper presents the practical application for a real-time and image-based driver assistance system. Dual-camera stereo vision algorithm for the distance measurement for the detected vehicle and multi-line merge function for the lane detection are implemented via lab-made library and calibrated off-line. A lab-made library developed via programming language is written and compiled for low computing-power consumption and rapidly calculating for the real-time purpose. To save the computing power consumption, some proper region of interest of vehicle detection and lane detection are applied. Finally, this system has been preliminary implemented on a test vehicle and driven on various roads over tens of miles and reach around 4 similar to 5 frame per seconds via portable embedded system.