基于机器视觉的车辆距离测量系统设计  

Design of Vehicle Distance Measurement System Based on Machine Vision

作　　者：唐玲[1] 陈明举[1] 杨平先[1] 

机构地区：[1]四川理工学院自动化与电子信息学院,四川自贡643000

出　　处：《实验室研究与探索》2016年第3期56-59,69,共5页Research and Exploration In Laboratory

基　　金：四川省教育厅科研项目(13ZB0138);人工智能四川省重点实验室开放基金项目(2013RYY02);四川理工学院校级科研基金项目(2012KY13)

摘　　要：为实现车距测量实时性与准确性的要求,机器视觉被引入到车距测量中,设计了具有实用性的单目视觉的车辆距离测量系统。该系统首先对采集到的图像利用小波变换进行去噪与增强,再通过Hough与模糊集论对车道线进行提取作为定位车辆的参考线以减少车辆定位的搜索时间,通过路面灰度值与基于窗口能量的方法准确定位车辆,最后根据针孔成像原理建立实际三维图像空间与摄像机二维图像的映射关系,计算出前面车辆的实际距离值。结果表明,当实际车辆距离小于70 m时,测距绝对误差小于2 m,相对误差不超过5%,测距时间小于0.25 s。In order to realize real-time and accurate measurement of vehicle distance system,the paper presents a prototype vehicle distance measurement system based on the monocular vision. This system first uses wavelet transformation to denoise and enhance the image,then extracts the lane line as reference of vehicle by using Hough and fuzzy set theory to reduce the searching time,uses grey value of the road and winder energy to locate the vehicle.Finally,by the geometric transformation theory mapping relationship between the actual 3D space and the 2D image space is established,and the leading vehicle distance is calculated. The experiment results show the absolute distance error is less than 2 meters provided that the actual distance is less than 70 m,the relative error is less than 5%,and the processing time is less than 0. 25 seconds. So this vehicle distance measurement system is real-time and accurate enough to utility.

关 键 词：机器视觉 车道线检测 图像去噪 距离测量 

分 类 号：TN911.73[电子电信—通信与信息系统]