but_calibration_camera_velodyne

ROS packages for extrinsic calibration Velodyne 3D LIDARs(vlp16) with ZED stereo camera. The package follows the methods proposed by:

• [1] Velas, Spanel, Materna, Herout: Calibration of RGB Camera with Velodyne LiDAR

• [2] Jingfeng Sui, Shuo Wang, Extrinsic Calibration of Camera and 3D Laser Sensor System. pdf

Requirements

• PCl 1.7 and OpenCV 3.x library

Usage

• roslaunch but_calibration_camera_velodyne calibration_coarse.launch

  • command will run the coarse calibration using 3D marker (described in the [1])
  • if the 3D marker detection fails, the attempt will be repeated after 5s
  • upon the cuccess the node ends and prints the 6 Degrees of Freedom of the Velodyne related to the camera

• roslaunch but_calibration_camera_velodyne calibration_fine.launch

  • will runs the calibration with the refinement. This process will take a longer time (approx. 1 minute)

• roslaunch but_calibration_camera_velodyne coloring.launch

  • launches the publisher of colored pointclouds created by the camera -- Velodyne fusion
  • the calibration parameters (6DoF) computed by the calibration process must be set in configuration file conf/coloring.yaml (see below)

Configuration:

Configuration *.yaml files are stored in the direcotry conf/:

• conf/calibration.yaml contains folowing parameters:
  • camera_frame_topic: name of the topic, where camera RGB image is published
  • camera_info_topic: topic of CameraInfo type
  • velodyne_topic: topic, where the Velodyne pointcloud is published
  • marker: physical size of the 3D marker
    • circles_distance: distance between the circles centres (see [1])
    • circles_radius: radius of the circles in marker (also see [1])
• in conf/calibration.yaml you can specify:
  • 6DoF: vector of the calibration parameters computed by the calibration process
  • velodyne_color_topic: name of the topic, where the colored pointcloud will be published

Sensors Setup:

Calibration Target:

Original Images:

2D edges:

After inverse distanse transform:

Circles detected on the edges image:

Original Pointcloud(front view and top view):

3D edges:

3d edges point reproject to image: