Awesome

hdl_people_tracking

hdl_people_tracking is a ROS package for real-time people tracking using a 3D LIDAR. It first performs <cite>Haselich's clustering technique</cite> to detect human candidate clusters, and then applies <cite>Kidono's person classifier</cite> to eliminate false detections. The detected clusters are tracked by using Kalman filter with a contant velocity model.

Video:<br>

Requirements

hdl_people_tracking requires the following libraries:

• OpenMP
• PCL 1.7

The following ros packages are required:

• pcl_ros
• <a href="https://github.com/koide3/ndt_omp">ndt_omp</a>
• <a href="https://github.com/koide3/hdl_localization">hdl_localization</a>

Example

Bag file (recorded in an outdoor environment):

• hdl_400.bag.tar.gz (933MB)

rosparam set use_sim_time true
roslaunch hdl_people_tracking hdl_people_tracking.launch

roscd hdl_localization/rviz
rviz -d hdl_localization.rviz

rosbag play --clock hdl_400.bag

[NOTE]:

If it doesn't work well, change ndt_neighbor_search_method in hdl_localization.launch to "DIRECT1". It makes the scan matching significantly fast, but a little bit unstable.

If your bagfile is static (velodyne device is fixed) try with the following launch file without any localization needs:

rosparam set use_sim_time true
roslaunch hdl_people_tracking hdl_people_tracking_static.launch

Related packages

• interactive_slam
• <a href="https://github.com/koide3/hdl_graph_slam">hdl_graph_slam</a>
• <a href="https://github.com/koide3/hdl_localization">hdl_localization</a>
• <a href="https://github.com/koide3/hdl_people_tracking">hdl_people_tracking</a>

Papers

Kenji Koide, Jun Miura, and Emanuele Menegatti, A Portable 3D LIDAR-based System for Long-term and Wide-area People Behavior Measurement, Advanced Robotic Systems, 2019 [link].

Contact

Kenji Koide, k.koide@aist.go.jp

Active Intelligent Systems Laboratory, Toyohashi University of Technology, Japan [URL]
Robot Innovation Research Center, National Institute of Advanced Industrial Science and Technology, Japan [URL]