A Method for Extrinsic Parameter Calibration of Rotating Binocular Stereo Vision Using a Single Feature Point

Area of Science:

• Computer Vision
• Robotics
• 3D Reconstruction

Background:

• Binocular stereo vision (BSV) is crucial for real-time 3D reconstruction.
• Current camera self-calibration methods often rely on time-consuming iterative optimization.
• Efficient self-calibration is needed for dynamic BSV applications.

Purpose of the Study:

• To develop a novel, rapid self-calibration method for BSV systems with rotating cameras.
• To overcome the limitations of iterative optimization in camera self-calibration.
• To improve the efficiency of 3D reconstruction in dynamic environments.

Main Methods:

• Established a BSV system with horizontally and vertically rotating, non-zooming cameras.
• Utilized Zhang's calibration method for initial intrinsic parameters and position estimation.
• Developed a new self-calibration technique using a single feature point and a quadratic equation for pitch and yaw angles.
• Derived closed-form solutions for pitch and yaw, avoiding iterative convergence issues.

Main Results:

• The proposed method achieves closed-form solutions for camera rotation angles (pitch and yaw).
• Computer simulations and physical experiments validated the method's feasibility.
• Experimental results showed average absolute errors of <0.21° for Euler angles and <6.6 mm for translation vectors.
• Average relative errors for 3D reconstruction coordinates did not exceed 4.2%.

Conclusions:

• The novel self-calibration method significantly reduces calibration time compared to iterative approaches.
• The method provides accurate estimation of camera parameters for rotating BSV systems.
• This technique enhances the efficiency and applicability of real-time 3D reconstruction.