ram-lab/lidar_appearance_calibration

Overview

This a ros package for multi lidar calibration by improving Qinghai's previous appearance-based work. Several method are to be implemented:

Dependency List

• PCL 1.8 (http://pointclouds.org/)
• Eigen 3
• Boost
• libpointmatcher (https://github.com/ethz-asl/libpointmatcher)
• Ceres-solver (http://ceres-solver.org/)
• YAML (https://github.com/jbeder/yaml-cpp)

This package was tested on Ubuntu 16.04, ROS Kinetic

Usage

1. Create a yaml file cfg.yaml into a fold, please follow ../data/example/top_tail/cfg.yaml to write
2. Preproces raw pointcloud to keep points in planar surfaces. You can use the below function or CloudCompare software to filter redundant points:
   
   • rosrun lidar_appearance_calibration calib_preprocess ../data/example/raw/ref.pcd ../data/example/raw/data.pcd ../data/example/raw/ref_filter.pcd ../data/example/raw/data_filter.pcd
   • 
3. Extract planes from pointcloud using RANSAC
   
   • rosrun lidar_appearance_calibration calib_plane_extraction pcd ../data/example/top_front/cfg.yaml
   • rostopic pub /contact/icp std_msgs/String "data: ''"
   • rviz -d ../rviz/plane_extraction
4. Visualize and check the extracted plane order (same colors mean that data are associated)
   • rosrun pcl_ros pcd_to_pointcloud ../data/example/top_front/plane/ref_planes.pcd 1
   • 
   • rosrun pcl_ros pcd_to_pointcloud ../data/example/top_front/plane/data_planes.pcd 1
   • 
   • rviz -d ../rviz/plane_extraction
   
   
5. Implement ICP to minimize Plane-to-Plane error
   • Auto initialization:
     rosrun lidar_appearance_calibration calib_icp ../data/example/top_front/ref_cfg.yaml ../data/example/top_front/data_cfg.yaml a
   • Manual initialization:
     rosrun lidar_appearance_calibration calib_icp ../data/example/top_front/ref_cfg.yaml ../data/example/top_front/data_cfg.yaml m
   • Call the program:
     rostopic pub /contact/save_plane std_msgs/String "data: ''"
6. Visualize the calibration result
   pcl_viewer ../data/example/top_front/merged_opt.pcd

Result

1. Plane extraction
   
   
2. Pointcloud fusion
   
   
3. Calibration resut
   
   

Reference

To use the code, pleace cite this paper:

@inproceedings{jiao2019novel,
  title={A novel dual-lidar calibration algorithm using planar surfaces},
  author={Jiao, Jianhao and Liao, Qinghai and Zhu, Yilong and Liu, Tianyu and Yu, Yang and Fan, Rui and Wang, Lujia and Liu, Ming},
  booktitle={2019 IEEE Intelligent Vehicles Symposium (IV)},
  pages={1499--1504},
  year={2019},
  organization={IEEE}
}