Extrinsic Camera Calibration with Line-Laser Projection.

Area of Science:

• Computer Vision
• Robotics
• Photogrammetry

Background:

• Precise camera pose estimation is crucial for multi-camera systems, with extrinsic calibration typically requiring in-situ marker-based methods.
• Existing methods often necessitate significant overlap between camera fields of view, limiting their applicability in certain fixed multi-camera setups.

Purpose of the Study:

• To develop a robust extrinsic calibration method for fixed multi-camera systems that functions effectively with little or no camera overlap.
• To provide a more practical and automatable solution for accurate camera pose determination in challenging calibration scenarios.

Main Methods:

• Projecting laser lines onto a planar surface (e.g., floor, wall) using a laser line projector.
• Optimizing camera and plane poses via bundle adjustment to align projected laser lines observed across cameras.
• Utilizing partial overlap between laser lines and camera fields of view for extrinsic calibration.

Main Results:

• Achieved rotation errors below 0.5° in real-world experiments, demonstrating high accuracy.
• Validated the method's effectiveness in scenarios with both overlapping and non-overlapping camera fields of view.
• Demonstrated comparable accuracy to state-of-the-art methods with a more practical calibration procedure.

Conclusions:

• The proposed laser line projection method offers a practical and accurate solution for extrinsic camera calibration in multi-camera systems, especially under limited overlap conditions.
• The technique is suitable for large-scale applications and can be fully automated, enhancing its utility in diverse robotic and vision tasks.