基于双目立体视觉技术的汽车测距系统实现

摘要

Automobile ranging is one of the key technologies of automobile anti-collision system.Based on analysing current common methods of automobile ranging,and aiming at their problems of complex equipment and serious mutual interference,we studied the application of machine vision in automobile ranging area and put forward a set of automobile ranging schemes,they are based on binocular stereo vision technology.Moreover,we employed OpenCV2.4.8 configuration in VS2012 compiler environment and used C /C ++programming to verify its feasibility,through the research on camera calibration,stereo correcting,stereo matching and other key technologies,we chose a fast stereo matching algorithm to measure the distance of targets.Experiment verified that the system has high feasibility and low cost,it only needs two ordinary CMOS cameras to complete image acquisition,and the platform is convenient in building, the ranging algorithm is easy to implement with fast speed and good scalability as well,and the accuracy within a certain scope meets the application requirements basically.%汽车测距是汽车防撞系统的关键技术之一。在分析现有的汽车测距常用方法基础上，针对常用测距方法设备庞杂、相互干扰严重等问题，对机器视觉在汽车测距方面的应用进行研究，提出一套基于双目立体视觉技术的汽车测距方案。并使用OpenCV2．4．8在 VS2012编译环境下用 C ／C ＋＋编程实现验证其可行性，通过对摄像机标定、立体校正、立体匹配等关键技术的研究，选用一种快速的立体匹配算法对目标进行测距。经过实验验证，该系统可行性高、成本低，只需两个普通 CMOS 摄像头便可完成图像采集工作，而且平台搭建方便，测距算法容易实现，速度快、扩展性好，在一定范围内的准确度基本满足应用要求。