Abstract In the non-contact detection of thread profile boundary correction, it remains challenging to ensure that the thread axis intersects the CCD camera axis perpendicularly. Here, we addressed this issue using modified algorithms. We established the Cartesian coordinate system according to the spatial geometric relationship of the thread. We used the center of the bottom of the thread as the origin, and the image of the extreme position image was replaced by the image of the approximate extreme position. In addition, we analyzed the relationship between the boundary of the theoretical thread image and the theoretical profile. We calculated the coordinate transformation of the point on the theoretical tooth profile and the coordinate function of the point on the boundary of the theoretical image. At the same time, the extreme value of the function was obtained, and the boundary equation of the theoretical thread image was deduced. The difference equation between the two functions was used to correct the boundary point of the actual thread image, and the fitting results were used to detect the key parameters of the external thread of the cylinder. Further experiment proves that the above algorithm effectively improves the detection accuracy of thread quality, and the detection error of main geometric parameters is reduced by more than 50 %.

中文翻译：

---

基于机器视觉的圆柱外螺纹牙型修正算法研究

摘要 在螺纹轮廓边界校正的非接触检测中，保证螺纹轴与CCD相机轴垂直相交仍然具有挑战性。在这里，我们使用修改后的算法解决了这个问题。我们根据螺纹的空间几何关系建立了笛卡尔坐标系。我们以螺纹底部的中心为原点，将极值位置图像替换为近似极值位置图像。此外，我们分析了理论螺纹图像的边界与理论轮廓之间的关系。我们计算了理论齿廓上点的坐标变换和理论图像边界上点的坐标函数。同时，得到函数的极值，推导出理论螺纹图像的边界方程。利用两个函数的差分方程对实际螺纹图像的边界点进行修正，并利用拟合结果检测圆柱外螺纹的关键参数。进一步实验证明，上述算法有效提高了螺纹质量的检测精度，主要几何参数的检测误差降低了50%以上。