dynamic_object_detection

Ros implementation of "Dynamic Object Detection in Range data using Spatiotemporal Normals" (published at the Australasian Conference on Robotics and Automation (ACRA) and available here).

Predictions (left) versus ground truth (right):

Paper:

If you find this work useful for your research, please consider citing our paper (submitted to ACRA 2023):

@inproceedings{falque2023dynamic,
    title={Dynamic Object Detection in Range data using Spatiotemporal Normals}, 
    author={Raphael Falque and Cedric Le Gentil and Fouad Sukkar},
    booktitle={Australasian Conference on Robotics and Automation, ACRA},
    year={2023}
}

Datasets:

Note

Some of the preprocessed ros bags are available here (hauptgebaeude folder for the DOALS dataset and HRI folder for the data from the UR5).

The method has been tested on a dataset collected with a UR5 robot arm.

We have also tested our approach on the undistorded scans from the Urban Dynamic Objects LiDAR Dataset (DOALS) (project page, direct link to downloads). See images above for samples from the dataset.

Requirements

Install ROS following these instructions (ROS2 is not implemented).

sudo apt install build-essential cmake libeigen3-dev libomp-dev git

To build:

git clone [email protected]:UTS-RI/dynamic_object_detection.git
ln -s ./dynamic_object_detection ~/catkin_ws/src
cd ~/catkin_ws
catkin_make

To run:

For the DOALS dataset:

roscore
rosbag play sequence_1.bag # undistorded LIDAR scans
roslaunch dynamic_detector normal_filter.launch

For the UR5 dataset:

roscore
rosbag play object_move_with_poses.bag
roslaunch dynamic_detector normal_filter_arm.launch