Enhancing Camera Calibration Accuracy With a Combination of Generalized Gaussian Model and Improved Warp Function

Camera calibration serves as a foundational step that directly influences the performance of vision-based measurement method. Among various target-based methods, circular patterns have been widely used due to their high detection accuracy and robustness. However, the process of feature extraction from circle patterns, being susceptible to eccentric errors during the circle projection transformation, is also affected by the interference of edge blurring and image noise. This study reports a novel approach that combines the generalized Gaussian model and the improved warp function. Firstly, the generalized Gaussian model is utilized to depict the image gradient distributions in the range of elliptical edge blurring, which is conducive to the precise coordinate calculation of the ellipse center. Furthermore, the improved warp function, which describes the projection transformation (including translation, scaling, and shearing) of the circle pattern, is designed to eliminate the projection errors. Simulations indicate that, compared with existing methods, the proposed method has stronger anti-noise and anti-blur interference capabilities. Experiments show that in the practical application of the binocular measurement system, the proposed method yields remarkable results and improves the measurement ability.