基于机器视觉的轻量级驾驶辅助系统

Machine Vision-based Lightweight Driver Assistance System

为了解决计算机视觉应用中数据量大、算法复杂的问题,根据道路结构特征和车辆行为特征,采用单个摄像头作为传感器,实现了一种轻量级的安全辅助驾驶系统。首先采用改进的边缘提取算法和车道线检测算法对摄像机内外参数进行离线标定;接着根据标定结果在二维平面图像上采用标识出实际空间距离的多窗口划分方法,并按不同的车间距将不同窗口划分为不同安全系数的区域,以赋予道路视觉检测的几何先验知识;当区域中出现障碍物时发出相应警示信息进行安全驾驶辅助,能为智能辅助驾驶提供轻量级的视觉检测平台。以便携式计算机和固定在车内的摄像头作为实验装置,在城市道路上进行车载实验。系统在车载实验中能够快速地提取车辆两侧的车道线,并利用离线标定的结果快速生成不同安全系数的警示区域,其中车辆在车道内正常行驶时的误检率和漏检率很小,可以忽略不计。与传统的驾驶辅助系统相比,本系统计算量大大降低,检测流程得到简化,可实现轻量级的车道和车辆检测,为系统在嵌入式系统上的实现奠定基础。

This paper proposed a machine vision-based lightweight driver assistance system. Firstly, the adiusted algorithm for extracting edge and lane line detection algorithm are used to calibrate inside and outside parameters of cameras offline. Secondly, a multi-window division method identifying the actual distance is used on two-dimensional image according to the results of calibration, and different window is divided into regions of different safety factor according to distance, in order to provide prior knowledge of geometry of vision detection to the road. Thirdly, when there is an obstacle in the area, the corresponding warning message is displayed to assist the driver and provide lightweight visual detection platform for intelligent driver assistance system. The proposed system in this paper can extract lane line on both sides of the vehicle quickly in car-board experiments and take advantage of off-line calibration results to generate alerts regions of different safety factors quickly,and both positive false detection rate and negative false detection rate in the experiment during normal driving in the lane are small and negligible. Compared with conventional driver assistance systems, our proposed method reduces the computation amount by simplifying the detection process to achieve lightweight lane and vehicle detection, and lays the foundation for implementation of the system on embedded systems.