Precise Calibration of Binocular Vision System Based on Oblique Cone Projection Model

The calibration accuracy of the binocular vision system is critical to its measurement accuracy. The existing methods are typically utilized to improve the calibration accuracy by nonlinear optimization of the previously obtained binocular system parameters in 2-D or 3-D space. Circular dot array targets are widely used in calibration due to their excellent noise immunity. However, the projection ellipse center is not the projection of the circular marker center. The eccentricity error will be substantial, especially with a large dip angle and close projection, impacting the calibration accuracy. Therefore, this article establishes a mathematical model for feature extraction error of circular dot array targets, analyzes the factors influencing the deviation between the projected ellipse center and the circle center projection, and evaluates the impact of center extraction error on the accuracy of the camera parameters solution. A new method for accurate calibration of the binocular vision system is also proposed. Based on the oblique cone projection model (OCPM), the 3-D points can be calculated by solving the circular cross sections and the central axes, where the cross Section circle centers are placed. Then, the parameters of the binocular system are optimized by minimizing the metric distance error between the reconstructed and real points in the 3-D measurement coordinate system and combining it with the constant distance constraint. The simulation and experimental results show that the calibration accuracy of this method is superior to the traditional algorithm and has potential application prospects.