gTRAVEL+: Enhancing Ground Segmentation for 3D Point Cloud with Temporal Noise Removal and Adaptive Plane Fitting for Urban and Off-Road Environments

gTRAVEL+ is an improvement of TRAVEL that focuses exclusively on enhancing ground segmentation performance.

🔍 Overview

gTRAVEL+ introduces:

• Temporal Node-Wise Noise Removal (TNNR) to eliminate noise.
• Merge Node Plane Fitting (MNPF) to tackle the under-segmentation issue caused by inappropriate region size.
• Rejected Ground Node Revert (RNGR) to recover misclassified ground points.

⚙️ How to build gTRAVEL+

1. Dependencies
   You need to install all the dependencies of the baseline method TRAVEL.
2. Build
   To successfully build gTRAVEL+, you can follow the commands below:

$ mkdir -p catkin_ws/src
$ cd catkin_ws/src
$ git clone https://github.com/giangdao1402/gTRAVEL-plus
$ ../
$ catkin build  # or catkin_make

🏃 How to run gTRAVEL+

$ cd catkin_ws
$ source devel/setup.bash
$ roslaunch gtravelp run_rellis.launch data_dir:=/media/giang/SSD/rosbag/SematicKITTI/Rellis_3D_os1_cloud_node_kitti_bin/Rellis-3D/00000 rviz:=True
 # or run_kitti.launch with data_dir of kitti bin

📊 Results

This method achieves state-of-the-art segmentation performance on urban (SemanticKITTI) and off-road (Rellis-3D) datasets.

Citation

If you use our codes, please consider citing our paper:

@ARTICLE{10990224,
  author={Dao, Truong Giang and Tran, Dinh Tuan and Nguyen, Duc Manh and Lee, Joo-Ho and Nguyen, Anh Quang},
  journal={IEEE Access}, 
  title={gTRAVEL+: Enhancing Ground Segmentation for 3D Point Cloud With Temporal Noise Removal and Adaptive Plane Fitting for Urban and Off-Road Environments}, 
  year={2025},
  volume={13},
  number={},
  pages={82431-82440},
  doi={10.1109/ACCESS.2025.3567621}}

Acknowledgement

We would like to thank the authors of Patchwork, Patchwork++ and TRAVEL for their invaluable open-sourced code.