Skip to content

README.md

Latest commit

 

History

History
81 lines (52 loc) · 2.78 KB

README.md

File metadata and controls

81 lines (52 loc) · 2.78 KB
Raw

Data Preparation and Format Conversion

If you have prepared your data in the following format:

timestamp x y z qw qx qy qz

Set your filenames and run pose2mat.m to get .mat files.

ELSE......

1. Data Preparation

  1. Drive the vehicle in a "$\infty$" shaped trajectory.

  2. Record LiDAR, camera, INS and IMU topics.

    rosbag record /velodyne_points /usb_cam_left/image_raw/compressed /novatel_data/inspvax /imu/data
    rosbag record /velodyne_points /zed2/zed_node/right/image_rect_color/compressed /novatel_data/inspvax /imu/data

  3. If there is no timestamp in INS topic (e.g. Novatel), use IMU timestamp as INS timestamp. They are usually hardware-synchronized. Make sure they have the same number of messages.

2. Pose Estimation

2.1 LiDAR-INS Calibration

  1. Record pose estimates of a SLAM algorithm (e.g. A-LOAM) and INS pose output.

    rosbag record -O calib /aft_mapped_to_init /novatel_data/inspvax /imu/data

  2. Convert the recorded topics into .csv format using my fork of rosbag_to_csv.

  3. Convert .csv files to .mat files.

    • readLO: Convert LiDAR odometry file to .mat format.
    • readNovatel: Convert Novatel pose file to .mat format.
    • readIMU: Convert IMU pose file to .mat format.

2.2 Camera-INS Calibration

  1. We use COLMAP to estimate and export camera pose. The .txt pose output should be in the following format:

    timestamp x y z qw qx qy qz

    where timestamp is the original timestamp when the image data is recorded.

  2. Convert INS and IMU topics into .csv format using rosbag_to_csv. Here we use IMU timestamp as the original INS timestamp.

  3. Convert .csv files to .mat files.

    • readVO: Convert visual odometry file to .mat format.
    • readNovatel: Convert Novatel pose file to .mat format.
    • readIMU: Convert IMU pose file to .mat format.

2.3 FCPE Calibration

  1. Use the methods above to acquire poses.

  2. Convert .csv files to .mat files.

    • readPose_FCPE: Convert LiDAR, camera and INS pose files to .mat format.

3. Interface Expansion

If you are using a different SLAM algorithm or a different INS device. You may define your own I/O interface accordingly. pose2mat.m can be used as a template.

4. Reference

  1. A-LOAM
  2. rosbag_to_csv
  3. COLMAP
  4. MATLAB-GPS-Calculations
  5. Novatel Driver