Probability-Based LIDAR-Camera Calibration Considering Target Positions and Parameter Evaluation Using a Data Fusion...

Area of Science:

• Robotics
• Computer Vision
• Geomatics

Background:

• Data fusion of 3-D light detection and ranging (LIDAR) point clouds and camera images is crucial for autonomous mobile robots.
• Accurate extrinsic calibration of LIDAR-camera systems is essential for precise 3-D mapping and object classification.

Purpose of the Study:

• To investigate the impact of calibration target deployment on data fusion accuracy.
• To identify key factors for effective target placement in extrinsic calibration.
• To propose a novel method for evaluating extrinsic calibration parameters.

Main Methods:

• Investigated the relationship between calibration target positions and data fusion accuracy.
• Developed a probability-based method for robust estimation of camera external parameters.
• Introduced an evaluation metric using the color ratio of 3-D colored point cloud maps.

Main Results:

• The proposed target deployment strategy significantly improves the estimation of camera external parameters.
• The probability density analysis confirmed the effectiveness of the deployment method.
• Quantitative evaluation demonstrated superior data fusion accuracy compared to existing methods.

Conclusions:

• Optimized calibration target deployment is critical for high-accuracy LIDAR-camera extrinsic calibration.
• The developed evaluation method provides a reliable way to assess calibration quality.
• This work advances the precision of 3-D mapping for autonomous systems.