期刊论文

Multi-frame point cloud fusion algorithm based on IMU and dynamic target detection

中文

计算机应用研究

1001-3695

2021

38

07

2179-2182

10.19734/j.issn.1001-3695.2020.07.0265

本校为第一机构

针对16线激光雷达点云数据稀疏,而导致环境感知效果不佳的问题,提出了一种基于惯性测量单元和动态目标检测相结合的多帧点云数据融合算法。该算法利用惯性测量单元提供的位姿信息进行点云中静态物体部分的融合,利用动态目标检测完成运动物体部分的融合,既可以增快点云融合的速度,又可以缓解融合时运动物体点云偏移过大的问题,在节约成本的同时,达到有效地增大点云密度的目的。实验结果表明,该算法进行点云融合时具有良好的效果,在无人驾驶环境感知方面具有较高的应用价值。

For the problem of poor environmental perception effect due to the sparse point cloud data of 16-line LiDAR, this paper developed a multi-frame point cloud data fusion algorithm based on inertial measurement unit and dynamic target detection. The algorithm used the pose information provided by the inertial measurement unit to fuse the static object parts in the point cloud, and used dynamic target detection to complete the fusion of the dynamic object parts, which could not only speed up the point cloud fusion speed, but also reduced the offset of the moving points cloud. It could effectively ...