Improved Camera Calibration Method and Accuracy Analysis for Binocular Vision

The quality of the stereo camera calibration method governs the accuracy of the binocular vision and thus determines the precision of the 3D reconstruction. Spurred by that, in this paper, we propose an improved calibration method that increases the calibration accuracy by optimizing the captured corners. Our method is multi-staged, where first, we employ the Harris corner detector to identify the uncertainties of all image corners. Then, for each detected corner, we apply a linear optimization, and finally, we calibrate the intrinsic and extrinsic parameters of the binocular vision setup by employing a nonlinear optimization scheme. We confirm the validity of our technique by analyzing the factors influencing the camera calibration accuracy and confirm the experimental conditions. Ultimately, we evaluate the proposed method and demonstrate that the mean error of our binocular vision architecture is more appealing compared to current literature.