基于单目视觉车距测量方法综述

车距测量是高级驾驶辅助系统(ADAS)的关键技术,是前碰撞预警和自适应巡航等功能的基础,单目摄像头相比于毫米波雷达等传感器价格低廉,获取图像信息量丰富,在ADAS系统中应用广泛,很多学者对单目测距技术进行了深入研究。论文主要介绍了四种基于单目视觉的车距测量方法,详细阐述了几何关系法、成像模型法、数据回归建模法和逆透视变换法的原理及测距模型,通过分析基本测距方法的不足,逐步介绍其改进形式。最后给出了每一种方法的特点,并对发展趋势进行展望。

Review of Vehicle Distance Measurement Based on Monocular Vision

Vehicle distance measurement is the key technology of Advanced Driving Assistance Systems(ADAS),and it is the basis of functions such as front collision warning and adaptive cruise control.Compared with sensors such as millimeter wave radar,monocular cameras are cheaper and have abundant image information.Thus,monocular cameras are widely used in ADAS systems,and many scholars have conducted in-depth research on monocular ranging technology.The paper mainly introduces four vehicle distance measurement methods based on monocular vision,expounds in detail the principles and models of geometric relationship method,imaging model method,data regression modeling method and inverse perspective transformation method.By analyzing the shortcomings of the basic methods,their improved forms are introduced.Finally the characteristics of each method are given,and the development trend is prospected.