Extrinsic Calibration between Camera and LiDAR Sensors by Matching Multiple 3D Planes

Area of Science:

• Robotics and Sensor Fusion
• Computer Vision
• 3D Perception

Background:

• Accurate extrinsic calibration is crucial for multi-sensor systems.
• Existing methods can be complex or require specialized equipment.
• Integrating image cameras and 3D LiDAR sensors presents unique calibration challenges.

Purpose of the Study:

• To propose a simple and effective extrinsic calibration method for a multi-sensor system comprising six cameras and a 16-channel 3D LiDAR.
• To accurately determine the rotation and translation between camera and LiDAR coordinate systems.
• To validate the method's performance using simulation and real-world data.

Main Methods:

• Utilizing a planar chessboard target for calibration.
• Reprojecting 2D camera-detected chessboard corners to a 3D plane in the camera coordinate system.
• Fitting 3D LiDAR point cloud data of the chessboard to a 3D plane in the LiDAR coordinate system.
• Calculating rotation by aligning plane normal vectors.
• Estimating translation by minimizing the distance between projected points on corresponding planes.
• Refining parameters using all 3D chessboard points and the LiDAR plane.

Main Results:

• The proposed method successfully calibrates the extrinsic parameters between cameras and LiDAR.
• Quantitative error analysis demonstrates the accuracy of the calibration.
• Calibration consistency is validated through experiments with real test sequences.

Conclusions:

• The developed extrinsic calibration method is simple, effective, and accurate for camera-LiDAR systems.
• The approach leverages planar targets for robust spatial relationship estimation.
• This work contributes to reliable sensor fusion in autonomous systems.