基于立体视觉的2D占据网格地图构建  

Construction of 2D occupying grid map based on stereo vision

作　　者：王永祥 常青 黄良红 孙永明[3] WANG Yongxiang;CHANG Qing;HUANG Lianghong;SUN Yongming(School of Information and Computer,Taiyuan University of Technology,Jinzhong 030600,China;Shanxi Air Traffic Management Bureau of Civil Aviation,Taiyuan 030031,China;Shanxi Academy of Forest Sciences,Taiyuan 030000,China)

机构地区：[1]太原理工大学信息与计算机学院,山西晋中030600 [2]民航山西空管分局,山西太原030031 [3]山西省林业科学研究院,山西太原030000

出　　处：《电子设计工程》2021年第24期1-6,共6页Electronic Design Engineering

基　　金：国家自然科学基金资助项目(61828601);山西省自然科学基金资助项目(201801D121141)。

摘　　要：目前大多数视觉SLAM系统只能输出相机轨迹与稀疏点云地图,无法适用于机器人进行导航、避障任务。通过单目相机虽然可以构建网格地图,但没有尺度信息,RGBD相机虽然具有尺度信息,但是在室外受环境光影响比较大。该文在ORB-SLAM2算法的基础上,采用双目视觉计算相机位姿,DepthToScan通过立体视觉的深度图像得到深度信息,在此基础上采用逆传感器模型计算占据概率,并在ROS通用平台下结合占据网格地图,重新构建了一种可用于机器人路径规划、导航的2D占据网格地图算法。实验表明,该方案能清晰标识出环境中障碍物的位置,建图精度相比于RTAB-map提高了3%。Currently,most visual SLAM systems can only provide camera trajectories and sparse point cloud maps,which are not suitable for robot navigation and obstacle avoidance tasks.Although monocular cameras can build grid maps,there is no scale information,RGBD cameras have scale information,but it is greatly affected by ambient light outdoors.Based on ORB⁃SLAM2,a stereo binocular vision camera is used to obtain depth information through a combination of stereo vision,DepthToScan and occupancy grid map on the ROS platform.This system was developed to rebuild a 2D occupational grid map algorithm that can be applied to robot path planning and navigation.The experiments show that the algorithm could clearly identify the location of obstacles in the environment.Compared with RTAB⁃map,the mapping accuracy of this algorithm is improved by 3%.

关 键 词：立体视觉 ORB-SLAM2 逆传感器模型 占据网格地图 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]