交通运输系统工程与信息

交通运输系统工程与信息 ›› 2016, Vol. 16 ›› Issue (4): 79-87.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于双目立体视觉的倒车环境障碍物测量方法

刘昱岗1a，王卓君1a，王福景1a，张祖涛*1b，徐宏2   

  1. 1. 西南交通大学a. 交通运输与物流学院；b. 机械工程学院，成都610031； 2. 中国电子科技集团第三十二研究所，上海200030
  • 收稿日期:2015-12-01 修回日期:2016-05-16 出版日期:2016-08-25 发布日期:2016-08-26
  • 作者简介:刘昱岗（1978-），男，湖南株洲人，副教授，博士.
  • 基金资助:

    国家自然科学基金项目/National Natural Science Foundation of China（51175443, 61271341）.

Vehicle Reversing Obstacle Measurement Based on Binocular-camera Stereo Vision

LIU Yu-gang1a,WANG Zhuo-jun1a,WANG Fu-jing1a, ZHANG Zu-tao1b, XU Hong2   

  1. 1a. School of Transportation and Logistics; 1b. School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China; 2. Thirty-second Research Institute of China Electronic Science and Technology Group, Shanghai 200030, China
  • Received:2015-12-01 Revised:2016-05-16 Online:2016-08-25 Published:2016-08-26

PDF

被引次数

6

摘要：

倒车引发的交通事故是主要的城市交通安全隐患之一，本文针对倒车安全及 障碍物检测中传统机器视觉倒车图像存在失真，以及距离感知精度低的问题，提出了一 种基于双目立体视觉的倒车环境障碍物测量方法.首先根据双目标定理论获取摄像头内、 外参数并分析摄像头畸变情况，使用标定参数对双目图像进行校正，运用极线约束使双 目图像平行共面，通过双目图像视差和三角测量原理获取图像中各目标的实际坐标，利 用固定单一物体计算测量距离与实际距离间的误差，进行双目立体视觉倒车障碍物检测 测量有效性研究.实验结果表明：摄像机横坐标与实际测量距离基本吻合，图像校正结果 比较理想，图像共面且极线对齐,能有效检测出后方障碍物，并有效提高倒车环境感知能 力和倒车安全性能.

关键词: 智能交通, 双目视觉, 倒车环境, 障碍物测量

Abstract:

It is one of the important urban traffic hidden troubles that accidents caused by driver reversing. In order to solve the shortcomings of traditional vehicle reversing technology based on machine vision, such as image distortions and low accuracy of distance perception, a vehicle reversing obstacle detection and measurement is presented based on binocular-camera stereo vision. The internal and the external parameters can be obtained according to the binocular calibration theory, which can be used to analyze the condition of the camera distortion. Then, calibration parameters are used to correct the binocular image. A method named epipolar constraint is used to get a pair of images coplanar. After this, the actual coordinates of each object in the image can be acquired through parallax of binocular image and the triangulation principle. Finally, the effectiveness of the obstacle measurement based on binocular camera stereo vision is verified by calculating the error between the measured distance and the actual distance, which is accomplished through fixing a single object. The experimental results show that the camera x-coordinate value is almost consistent with the actual results. Stereo rectification removes distortions and turning the stereo pair of images into standard aligned form which are nearly coplanar. The method can detect the rear obstacles effectively, and improve the ability to reverse the environmental awareness. Besides, it can improve the safety performance of the reversing process.

Key words: intelligent transportation, binocular vision, vehicle reversing environment, obstacle measurement

中图分类号: