摘要
提出了一种基于单目视觉序列图像的车辆自运动估计算法,首先利用自车运动和相机运动间的约束建立了运动模型,该模型消除了运动参数间的模糊性从而提高了运动估计的可靠性;然后提出了基于多参考帧的运动估计实现技术,增强了算法在低速场景下的准确性。实验结果表明该算法不容易受光照和背景的影响,在复杂应用场景中能够准确的估计出车辆自运动参数。
A robust vehicle ego-motion estimation algorithm based on image sequence from a monocular camera was proposed.First,motion model was established based on the constraint between vehicle motion and camera motion.The new model eliminated the ambiguity between rotational and translational ego-motion parameters and improved the reliability of estimation result.Then,the ego-motion parameters were estimated based on a multiple reference frames technique.And it improved the accuracy of estimation result when vehicle was in low speed.The experiment results show that the presented method is less prone to the influence of illumination or backgrounds,and can estimate the parameters accurately in complex application environment.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2011年第3期552-557,共6页
Journal of System Simulation
基金
国家863计划项目(2006AA11Z221)
关键词
自运动估计
运动模型
鲁棒
单目视觉
辅助驾驶系统
ego-motion estimation
motion model
robustness
monocular vision
driver assistance system