Planar self-calibration for stereo cameras with radial distortion

Area of Science:

• Computer Vision
• Robotics
• Photogrammetry

Background:

• Stereo calibration is crucial for 3D reconstruction and depth perception.
• Existing methods often require multiple images, special hardware, or are sensitive to radial distortion.

Purpose of the Study:

• To develop a robust and efficient stereo calibration technique.
• To provide an exact algebraic solution using minimal correspondences.
• To handle significant radial distortions without special hardware.

Main Methods:

• Solving a polynomial equation system with two radial distortion parameters.
• Utilizing a minimal set of five image point correspondences.
• Employing Random Sample Consensus (RANSAC) for outlier rejection.
• Performing non-linear optimization for intrinsic and extrinsic parameters.
• Enabling joint optimization for multiple stereo image pairs.

Main Results:

• Achieved accurate stereo calibration from a single stereo image pair of a planar scene.
• Demonstrated robustness against severe radial distortions.
• Reduced computation time and increased robustness with minimal correspondences and RANSAC.
• Validated efficiency and accuracy on both synthetic and real-world data.

Conclusions:

• The proposed method offers an exact algebraic solution for stereo calibration.
• It is robust, efficient, and accurate, even with challenging image data.
• The technique eliminates the need for specialized hardware, making it widely applicable.