Front parallel alignment of circular patterns for improvement of OpenCV camera calibration

In this paper, we propose a method that combines perspective rectification with iterative feature point detection for circular grid patterns. A homography matrix estimated from initially detected points and their ideal coordinates is used to rectify the pattern image to a frontal view, restoring circles distorted into ellipses. Feature points are then re-detected on the rectified image and inverse-transformed to the original coordinates. Repeating this process improves detection accuracy. Simulations with varying rotation angles and lens distortion coefficients show that our method achieves greater robustness to projective distortion than OpenCV’s cvfindCirclesGrid.