Hand-eye and radial distortion calibration for rigid endoscopes

Area of Science:

• Robotics
• Computer Vision
• Calibration Techniques

Background:

• Traditional hand-eye calibration methods often separate intrinsic and extrinsic parameter estimation or assume a pinhole camera model.
• Existing approaches may not effectively address radial distortion in camera systems, impacting 3D reconstruction fidelity.

Purpose of the Study:

• To develop a non-linear calibration method for hand-eye systems that simultaneously estimates camera intrinsic parameters, including radial distortion, and the hand-eye transform.
• To improve the accuracy of 3D reconstruction in robotic systems using endoscopes with radial distortion.

Main Methods:

• The proposed method employs a three-step approach: initial linear estimation of parameters, expressing camera extrinsics relative to hand-eye and world-grid transforms, and bundle adjustment for simultaneous refinement.
• Utilizes a minimum of three frames of a calibrated pattern for initial estimation and refinement.
• Incorporates radial distortion parameters directly into the bundle adjustment process.

Main Results:

• The non-linear calibration method achieves a 3D reconstruction error of approximately 5% of the object's size.
• Quantitative comparisons demonstrate superior performance against state-of-the-art linear and non-linear calibration techniques.

Conclusions:

• Simultaneous estimation of hand-eye and radial distortion parameters significantly enhances 3D reconstruction accuracy.
• The developed method offers an effective solution for calibrating robotic systems with radially distorted camera endoscopes.