摘要
同时定位与地图构建SLAM是移动机器人、自动驾驶车等在未知环境实现自主导航的关键技术,针对目前激光雷达价格昂贵,为降低成本,该文采用RGB-D传感器,提出一种改进的RGB-D SLAM系统,进一步提升导航定位的精确性和高效性。主要步骤是获取图像与参数采集、特征提取与匹配、相机位姿估计、位姿图优化、闭环检测和地图构建与转换等,输入为RGB-D摄像机提供的彩色数据以及深度数据;输出为可供导航和避障的三维地图,目的是将RGB-D的深度数据转换为激光雷达数据并封装成消息进行广播。根据RGB-D姿态和闭环检测,得到稠密三维点云,并对点云地图利用八叉树方法得到一种用于机器人导航的三维地图。根据实验对比,结果表明所构建的RGB-D SLAM系统在精确性和高效性有良好效果。
Simultaneous positioning and map construction slam are the key technologies for mobile robots and autonomous vehicles to realize autonomous navigation in unknown environment.Aiming at the high price of lidar,in order to reduce the cost,this paper uses RGB-D sensor to realize slam navigation,and proposes an improved RGB-D slam system to further improve the accuracy and efficiency of navigation and positioning.The main steps are to obtain image and parameter acquisition,feature extraction and matching,camera pose estimation,pose map optimization,Closed-loop Detection and map construction and conversion;input color data and depth data provided by RGB-D camera,and output to 3D map for navigation and obstacle avoidance,with the purpose of converting RGB-D depth data into lidar data and encapsulating them into messages Broadcast.According to RGB-D attitude and closed-loop detection,dense 3D point cloud is obtained,and a 3D map for robot navigation is obtained by using octree method.According to the experimental comparison,the results show that the RGB-D slam system has good accuracy and efficiency.
作者
王增喜
张庆余
贾通
张苏林
WANG Zeng-xi;ZHANG Qing-yu;JIA Tong;ZHANG Su-lin(Automotive Data of China(Tianjin)Co.,Ltd.,Tianjin 300393,China)
出处
《自动化与仪表》
2021年第4期73-78,共6页
Automation & Instrumentation
基金
天津市科技支撑重点项目(18YFZCGX00390)。
