摘要
本文基于立体视觉定位技术,提出了基于双目立体视觉的栅格地图构建方法,用以解决目前视觉SLAM技术构建的稀疏特征地图难以直接用于自主导航的问题。本文提出的方法仅以视觉信息作为输入实时完成移动机器人自定位与外界环境栅格地图的构建。首先采用双目立体视觉定位获取机器人运动参数,利用稠密匹配估算空间点云分布,在考虑机器人实际高度的情况下将三维点云投影成二维数据,最后通过二值贝叶斯滤波器在线构建栅格地图。本文所构建的栅格地图包含环境几何信息,可直接应用于机器人路径规划与导航。实验结果验证了本文所以出的定位与地图构建方法的可行性。
To solve the problem that map obtained from SLAM cannot be directly used to do path planning and autonomous navigation, an approach based on stereovision to build grid map is proposed. The method completes location and map building with visual information as the only input. Visual odometer is used to estimate motion parameters of mobile robot between previous and current flame on real time; dense stereo matching is used to obtain 3D point cloud with dept information of environment; binary bayesian filter was used to build grid map when projection 3D points cloud to 2D information. The grid map can be used to do path planning and navigation directly for it includes geometry information of environment. The experimental results demonstrate the feasibility of the stereovision-based location and mapping approach.
出处
《软件》
2012年第11期233-236,共4页
Software
基金
国家自然科学基金自主项目(50775013)
高等学校科技创新工程重大项目培育项目资金资助(708011)
