【摘要】 为提高室内移动机器人对静态与动态障碍物的感知精度，本文提出一种多传感器点云数据融合算法。首先将深度相机数据转换为点云，并提出超声波数据点云化算法，将一维声波数据扩展为二维点云；随后融合激光雷达、相机及超声波点云，形成统一的环境表征。实验表明，相较于单一激光雷达，本算法对镂空与透明静态障碍物的检测相对误差分别降低94.64%与95.04%，动态障碍物检测误差减少84.32%，显著提升了环境感知的准确性。更多还原

【Abstract】 To improve the perception accuracy of indoor mobile robots towards static and dynamic obstacles, this paper proposes a multi-sensor point cloud data fusion algorithm. Firstly, the depth camera data is converted into a point cloud, and an ultrasonic data point cloud transformation algorithm is proposed to extend one-dimensional acoustic data into two-dimensional point clouds; Subsequently, laser radar, camera, and ultrasonic point cloud are integrated to form a unified environmental representation. Experiments have shown that compared to a single LiDAR, this algorithm reduces the relative error of detecting hollow and transparent static obstacles by 94.64% and 95.04%, respectively, and reduces the error of detecting dynamic obstacles by 84.32%, significantly improving the accuracy of environmental perception.更多还原