一种基于VSLAM的室内导航地图制备方法

摘要

针对室内复杂环境下导航地图制备问题,提出一种采用图像传感器进行导航地图制备的方法.首先,基于VSLAM技术估计相机位姿,经后端优化后生成环境三维点云;其次,筛选出地面点云,基于随机采样一致性算法得到精确的地面平面方程;最后,投影机器人通行环境区域的点云至二维栅格地图,通过利用贝叶斯方法更新栅格占用状态后,实现室内导航地图的制备.实验结果表明,基于VSLAM制备的室内导航地图精度满足导航要求,且不同于传统激光地图构建方法,地图中包含环境的三维结构信息,在进行路径规划实验中,避免了规划的路径穿过低矮物体及悬空物体的错误.%A method was proposed for navigation mapping under complex indoor environment by using the image sensor.Firstly,the pose of the camera was estimated based on VSLAM,and we got the 3D point cloud after the back-end optimization.Secondly,the ground point cloud was selected,and the ground plane equation was obtained based on the random sampling consensus algorithm.Finally,after projecting the point cloud of robots' passing area to the 2D grid map and updating the occupancy state of the grid by using Bayesian method,the indoor navigation mapping was implemented.Experimental results show that:1) The accuracy of indoor navigation map based on VSLAM can meet the navigation requirements,and 2) Different from the traditional mapping based on laser,the map based on VSLAM contains 3D environment structure information.In the path planning experiment,the planned path avoided the error of passing through the low objects and the suspended objects.