Novel Bayesian Inference-Based Approach for the Uncertainty Characterization of Zhang's Camera Calibration Method

Area of Science:

• Computer Vision
• Machine Learning
• Metrology

Background:

• Camera calibration is crucial for machine vision, with Zhang's method widely used for intrinsic parameters and lens distortion.
• Current methods for estimating parameter uncertainty may lack reliability, potentially introducing biases in post-processing.

Purpose of the Study:

• To develop and evaluate a novel Bayesian inference-based approach for camera calibration.
• To assess the certainty and reliability of intrinsic and extrinsic camera parameters compared to Zhang's method.

Main Methods:

• Utilized Bayesian inversion to recalibrate intrinsic camera parameters, assuming Zhang's estimates as the prior probability.
• Developed a new procedure for optimizing extrinsic parameters within the Bayesian framework.

Main Results:

• Bayesian recalibration yielded different uncertainties for intrinsic parameters compared to Zhang's method.
• The primary source of inaccuracy was identified in the extrinsic parameter calculation procedure.
• The novel Bayesian approach significantly improved the reliability of image point predictions by optimizing extrinsic parameters.

Conclusions:

• The proposed Bayesian inference method enhances the reliability of camera calibration, particularly in extrinsic parameter estimation.
• This approach offers improved accuracy for machine vision applications requiring precise camera parameter estimation.