Approach for accurate calibration of RGB-D cameras using spheres.

Area of Science:

• Computer Vision
• Robotics
• Sensor Fusion

Background:

• RGB-D cameras are crucial for many vision tasks.
• Standard calibration methods lack the required accuracy for precise applications.
• Accurate intrinsic and extrinsic parameters are essential for reliable sensor data.

Purpose of the Study:

• To propose a novel and accurate sphere-based calibration framework for RGB-D cameras.
• To develop a depth error correction method for improved performance.
• To enhance the precision of intrinsic and extrinsic parameter estimation.

Main Methods:

• A sphere-based calibration framework utilizing feature extraction to detect sphere projections.
• Automatic detection of sphere center and edges, excluding noise and outliers.
• Closed-form solution for initial parameters using sphere center projections on RGB and depth images.
• Nonlinear global minimization for accurate final parameter estimation.
• Detailed analysis of the depth error correction principle.

Main Results:

• The proposed method achieves higher calibration accuracy compared to state-of-the-art techniques.
• The feature extraction module reliably identifies sphere features, ensuring robustness.
• The framework provides an easy-to-use solution for accurate RGB-D camera calibration.
• Experimental analysis validates the effectiveness and accuracy of the proposed method.

Conclusions:

• The sphere-based calibration framework offers a significant improvement in accuracy for RGB-D cameras.
• The integrated depth error correction enhances the overall reliability of the sensor pair.
• This method is a practical and effective solution for demanding vision applications requiring precise calibration.