基于单目视觉的自车运动参数鲁棒估计  

A Robust Method for Vehicle Ego-Motion Estimation Based on Monocular Vision

作　　者：刘威[1] 魏存伟[1] 赵逢[1] 段勃勃[1] 袁淮[1] 赵宏[1] 

机构地区：[1]东北大学软件中心,辽宁沈阳110179

出　　处：《机器人》2009年第1期20-24,32,共6页Robot

摘　　要：提出一种适用于结构化道路的自车运动参数鲁棒计算方法.该方法通过建立双线性增量自运动模型,选择车道线和道路导向箭头上的角点作为路面特征点,利用连续两幅图像的多对匹配的特征点对模型进行求解,从而估计出自车运动参数.实验结果表明该方法不容易受光照和背景的影响,在结构化道路上能够准确地估计自车运动参数.同时该方法计算量小,能够满足实时性要求.A robust estimation method of ego-motion parameters is proposed for structural road. First, bilinear incremental ego-motion model is established, and the corners of lane markers and road markings are extracted as feature points. Then, the motion model is solved with the matched comers in the two sequential images, and the ego-motion parameters are estimated. The experiment results show that the presented method is less prone to the influence of illumination or backgrounds, and can estimate the parameters accurately on the structural road. Moreover, the proposed method is simple enough to meet the real-time requirements.

关 键 词：自车运动估计 结构化道路 双线性增量模型 角点检测和匹配 单目视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]