基于ORB-SLAM的室内机器人定位和三维稠密地图构建

发布时间：2018-05-03 21:27

  本文选题：同时定位和地图构建 + 室内机器人　； 参考：《计算机应用》2017年05期

【摘要】：针对在室内机器人定位和三维稠密地图构建系统中,现有方法无法同时满足高精度定位、大范围和快速性要求的问题,应用具有跟踪、地图构建和重定位三平行线程的ORB-SLAM算法估计机器人三维位姿;然后拼接深度摄像头KINECT获得的三维稠密点云,提出空间域上的关键帧提取方法剔除冗余的视频帧;接着提出子地图法进一步减少地图构建的时间,最终提高算法的整体速度。实验结果表明,所提系统能够在大范围环境中准确定位机器人位置,在运动轨迹为50 m的大范围中,机器人的均方根误差为1.04 m,即误差为2%,同时整体速度为11帧/秒,其中定位速度达到17帧/秒,可以满足室内机器人定位和三维稠密地图构建的精度、大范围和快速性的要求。
[Abstract]:In order to solve the problem that the existing methods can not meet the requirements of high accuracy, large range and fast in the indoor robot localization and 3D dense map construction system, the application has tracking. The ORB-SLAM algorithm of map construction and resetting three parallel threads is used to estimate the position and pose of the robot, and then the 3D dense point cloud obtained by the depth camera KINECT is spliced together, and the key frame extraction method in spatial domain is proposed to remove redundant video frames. Then submap method is proposed to further reduce the time of map construction, and finally improve the overall speed of the algorithm. The experimental results show that the proposed system can accurately locate the position of the robot in a wide range of environments. In a large range of motion tracks of 50 m, the root-mean-square error of the robot is 1.04 m, that is, the error is 2 parts, and the overall speed is 11 frames / s. The speed of localization can reach 17 frames / s, which can meet the requirements of accuracy, wide range and rapidity of indoor robot localization and 3D dense map construction.
【作者单位】： 华南理工大学自动化科学与工程学院;
【基金】：国家自然科学基金资助项目(61573148) 广东省科技重大专项(2015B010919007)~~
【分类号】：TP242
，

本文编号：1840173