A simple and precise calibration method for binocular vision

Abstract

Binocular vision is an important part of machine vision measurement. Calibration accuracy is crucial for binocular vision. As for the determination of the structure parameters of the two cameras, the existing approaches usually obtain the initial values and optimize them according to the image-space errors, object-space errors or a combination of these. In the optimization process, constructing the objective function only through the image-space errors or object-space errors is not enough. Moreover, the image-space and object-space errors can form a variety of combinations to construct the objective function. Therefore, it is hard to choose the error criterion. An inadequate error criterion may lead to over-optimized or local minima (ambiguity solution). To solve this problem, this paper proposes a simple and precise calibration method for binocular vision based on the points distance constraints and image-space errors. The process of determining the structure parameters was divided into noniterative and iterative parts. We calculated the structure parameters of the two cameras according to the distance constraints of every two feature points noniteratively. The results obtained in this step were set as the initial value and refined through minimizing the reprojection errors using the Levenberg–Marquardt method. Because the results obtained in the noniterative step are accurate enough, only one iteration is needed. In this way, we finish the calibration avoiding the need to choose the error criterion. Furthermore, our method reduces the number of iterations to improve the calibration efficiency on the premise of guaranteeing the calibration accuracy. The experimental results show the superiority of this calibration method compared with other calibration methods. Using the calibration results of our method, in the measurement range of −45°∼ 45°, the rotation angle measurement error was less than ±0.032°. In the measurement range of 0 ∼ 39 mm, the displacement measurement error was less than ±0.047 mm. As for the length measurement of a 300 × 225 mm target, the length measurement error was less than ±0.039 mm.