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.

中文翻译：

---

改进的双目视觉相机标定方法及精度分析

立体相机校准方法的质量决定了双目视觉的准确性，从而决定了 3D 重建的精度。受此启发，在本文中，我们提出了一种改进的校准方法，通过优化捕获的角来提高校准精度。我们的方法是多阶段的，首先，我们使用 Harris 角点检测器来识别所有图像角点的不确定性。然后，对于每个检测到的角，我们应用线性优化，最后，我们通过采用非线性优化方案校准双目视觉设置的内在和外在参数。我们通过分析影响相机校准精度的因素并确认实验条件来确认我们技术的有效性。最终，