This ROS package reads pointcloud data from 5 LiDARs and estimates the position and some attitude information of the clusters.
All inputs are in type sensor_msgs/PointCloud2.
/livox/lidar_192_168_147_230/livox/lidar_192_168_147_231/livox/lidar_192_168_147_232/livox/lidar_192_168_147_233/livox/lidar_192_168_147_234
Outputs are in topics:
/filter/lidarinsensor_msgs/PointCloud2: This is the processed pointcloud./filter/targetingeometry_msgs/PoseArray: This is the position and some attitude information of the clusters. Each cluster will occupy twogeometry_msgs/Poseposes in the topic.- Therefore the length of messages in the topic = 2 * the number of clusters.
- For the nth (n = 1, 2, ..., n, ...) cluster, its information will be structured as:
- geometry_msgs/Pose[2*n-2]:
- position (related to the mass center of the USV)
- x: x of the cluster center
- y: y of the cluster center
- z: z of the cluster center
- orientation: x, y, z, w are the quaternions of the clusters, only containing the heading (yaw) info of the cluster.
- position (related to the mass center of the USV)
- geometry_msgs/Pose[2*n-1]:
- position (related to the mass center of the USV)
- x: x of the highest point of the cluster
- y: y of the highest point of the cluster
- z: z of the highest point of the cluster
- orientation:
- x: length of the cluster
- y: width of the cluster
- z & w are reserved
- position (related to the mass center of the USV)
- geometry_msgs/Pose[2*n-2]:
- Box filter is used for removing useless or meaningless points
- Outlier filter is used for removing noise points
- L-shape fitting is used for attitude estimation
- Search-based rectangle fitting is used to estimate position
- OpenMP is used for increasing process speed
- Parameters can be adjusted in launch files
- L-shape fitting from L-shape-fitting-3D-LiDAR-ROS is used and modified for faster process speed.
- Dr. Lv for providing the basic frame of the code.