High-precision method of binocular camera calibration with a distortion model.

Area of Science:

• Computer Vision
• Robotics
• Photogrammetry

Background:

• Accurate camera calibration is essential for 3D reconstruction in binocular stereo vision systems.
• Existing methods often face challenges with precision and efficiency.

Purpose of the Study:

• To develop a high-precision, cost-efficient, and simple camera calibration method for binocular stereo vision.
• To optimize intrinsic and extrinsic camera parameters, including lens distortion.

Main Methods:

• Utilized a multi-view calibration template with diverse encoded points.
• Employed alternative bundle adjustment for optimizing intrinsic and extrinsic parameters.
• Leveraged monocular calibration algorithms for initial parameter estimation.

Main Results:

• Achieved maximum mean absolute relative calibration errors of 3.5e-6 for focal length and 1.2e-6 for the principal point in simulations.
• Real-world data demonstrated a reprojection error of approximately 0.045 pixels.
• Obtained high precision for lens distortion, intrinsic, and extrinsic parameters.

Conclusions:

• The proposed method offers a convenient, cost-effective, and highly precise solution for binocular stereo vision calibration.
• The technique is robust and suitable for various applications requiring accurate 3D spatial information.