Tomato1107 / multi_lidar_calibration

Calibrate extrinsic parameters of multi-lidars, based on ICP or NDT, etc.

Multi Lidar Calibration

Usage

Step 1. segment significant parts of cloud

The original point cloud should be first segmented in order to eliminate the influences of unimportant features.

1. modify crop.launch file

<?xml version="1.0"?>
<launch>
    <group ns="lidar_0">
        <param name="mode" value="single"/>
        <param name="input_topic" value="/lidar_0/lidar_node/pointcloud"/> <!-- TODO -->
        <param name="output_topic" value="/lidar_0/pointcloud_filtered"/> <!-- TODO -->

        <node pkg="multi_lidar_calibration" type="cloud_cropper_node" name="cloud_cropper_node" output="screen"/>
        <node pkg="multi_lidar_calibration" type="cloud_crop_helper.py" name="cloud_crop_help" output="screen"/>
    </group>
</launch>

2. roslaunch multi_lidar_calibration crop.launch to find significant features.

3. write down the segment parameters in the configuration file.

input_topic: "/lidar_0/lidar_node/pointcloud"

output_topic: "/lidar_0/pointcloud_filtered"

boxes:
  box0:
    x: -2.69
    y: -0.50
    z: -0.81
    a: 0.19
    b: 0.86
    c: 0.55

  box1:
    x: -2.89
    y: 1.33
    z: -0.82
    a: 0.28
    b: 0.27
    c: 1.40

  box3:
    x: 2.35
    y: 0.18
    z: 0.56
    a: 0.52
    b: 0.56
    c: 0.37

Step 2. Run calibration algorithm

<?xml version="1.0"?>
<launch>
  <param name="use_sim_time" value="true"/>


  <!-- 1 - 0 -->
  <group ns="lidar_1">
      <param name="config" value="$(find multi_lidar_calibration)/config/crop/lidar_1.yaml"/> <!-- TODO -->
      <param name="mode" value="multi"/>

      <node pkg="multi_lidar_calibration" type="cloud_cropper_node" name="cloud_cropper_node" output="screen"/>
  </group>

  <group ns="lidar_0">
      <param name="config" value="$(find multi_lidar_calibration)/config/crop/lidar_0.yaml"/> <!-- TODO -->
      <param name="mode" value="multi"/>
      
      <node pkg="multi_lidar_calibration" type="cloud_cropper_node" name="cloud_cropper_node" output="screen"/>
  </group>


  <node pkg="multi_lidar_calibration" type="multi_lidar_calibration_node" name="multi_lidar_calibration_node" output="screen">
    <param name="parent_cloud_name" value="/lidar_0/pointcloud_filtered"/> <!-- TODO -->
    <param name="child_cloud_name" value="/lidar_1/pointcloud_filtered"/> <!-- TODO -->
    <param name="voxel_size" value="-1"/>

    <param name="method" value="ICP"/>
    <param name="config_file" value="/home/ubuntu/Data/workspace/github_projects/lidar_calibration/src/multi_lidar_calibration/config/icp.yaml"/>
  </node>
</launch>

1. roslaunch multi_lidar_calibration calibration.launch to start

The registration process is actually based on these point cloud:

And this is the whole point cloud:

Result