ImMesh_PGO: An Immediate LiDAR Localization and Meshing Framework with Loop Closure

Comparison of Consistent Mapping

(Left) Photo of the Scene, (Middle) ImMesh, (Right) ImMesh_PGO running on unsw_lowercampus_20240916.bag.

1. Prerequisites

1.1 ROS

Following this ROS Installation to install ROS and its additional pacakge:

sudo apt-get install ros-XXX-cv-bridge ros-XXX-tf ros-XXX-message-filters ros-XXX-image-transport ros-XXX-image-transport*

NOTICE: remember to replace "XXX" on above command as your ROS distributions, for example, if your use ROS-noetic, the command should be:

sudo apt-get install ros-noetic-cv-bridge ros-noetic-tf ros-noetic-message-filters ros-noetic-image-transport*

1.2. livox_ros_driver

Follow this livox_ros_driver Installation.

1.3 CGAL and OpenGL

sudo apt-get install -y libcgal-dev pcl-tools
sudo apt-get install -y libgl-dev libglm-dev libglfw3-dev libglew-dev libglw1-mesa-dev 
sudo apt-get install -y libcgal-dev libxkbcommon-x11-dev

1.4 GTSAM

Follow this GTSAM Get Started.

2. Build ImMesh on ROS

Clone this repository and catkin_make:

cd ~/catkin_ws/src
git clone https://github.com/YixFeng/ImMesh_PGO.git
cd ../
catkin_make
source ~/catkin_ws/devel/setup.bash

3. Run

Use this command, you can run unsw_lowercampus_20240916.bag.

roslaunch immesh_pgo mapping_horizon.launch
rosbag play unsw_lowercampus_20240916.bag

Since I only incrementally updated the mesh after the loop-closure optimization, the old mesh features have not been removed and will be mixed with the new ones.

If you want to have a map constructed from loop-closure-optimized poses, you can follow instructions below:

1. Run ImMesh_PGO.

roslaunch immesh_pgo mapping_horizon.launch

2. Click "Save PGO Cloud to file" in the UI and you will get a corrected pointcloud in your ~/ImMesh_PGO directory.

3. Build the mesh using the pointcloud. You need to change the path in the launch file.

roslaunch immesh_pgo mapping_pointcloud.launch

Acknowledgement

Thanks for following open-sourced repositories and authors:

• ImMesh: An Immediate LiDAR Localization and Meshing Framework
• STD: A Stable Triangle Descriptor for 3D place recognition
• FAST_LIO_SLAM: Integration of FAST-LIO2 and Scan Context-based Loop Detection