为有效解决单一传感器同时定位与地图构建(simultaneous localization and mapping,SLAM)定位精度低、障碍物识别不全问题,提出一种多传感器融合的SLAM方法.通过将RGB-D相机采集的点云进行降采样、滤波处理,极大降低算法的计算量.利用点云库对激光点云和降采样RGB-D相机点云进行融合,融合的点云利用PL-ICP完成点云配准,提高对外部环境的准确识别.利用扩展卡尔曼滤波融合IMU和轮式里程计与点云进行位姿匹配,保证定位的精度.实验结果表明,该方法可以有效提高对室内建图和导航的精度.
Abstract
To effectively solve the problems of low accuracy of single sensor SLAM and incomplete obstacle recognition,a multi-sensor fusion SLAM method was proposed.The computational effort of the algorithm was greatly reduced by down-sampling and filtering the point cloud acquired by the RGB-D camera.The laser point cloud and the down-sampling camera point cloud were fused using the point cloud library,and the point cloud alignment was accomplished using PL-ICP on the fused point cloud,which improved the accurate recognition of the external environment.Extended Kalman filtering was utilized to fuse the IMU and wheeled odometer with the point cloud for position matching,thus ensuring the accuracy of positioning.Experimental results show that the method can effectively improve the accuracy of indoor mapping and navigation.
关键词
多传感器融合/同时定位与地图构建/移动机器人/扩展卡尔曼滤波/惯性传感器/点云配准/RGB-D相机
Key words
multi-sensor fusion/SLAM/mobile robot/extended Kalman filter/inertial sensor/point cloud registration/RGB-D camera
维普
万方数据