“…At present, many relevant scholars at home and abroad have conducted research on this subject and have already formed a relatively complete thread detection system, as in [3]- [8], which can be described as follows: First, the original image of the thread profile by CCD or other optical camera is captured, then the features of thread profile based on image processing techniques such as filtering, boundary extraction and curve fitting are extracted, and finally the parameters are measured from the acquired features such as the thread profile, which are considered as the result of detection. Min and Zhao respectively designed a non-contact thread parameter measurement system based on machine vision to support efficient and accurate measurement of the thread contact angle, as in [9], [10]; Rao et al summarized the image processing technology and computer vision algorithms currently used for external thread detection, as in [11]; Senthilnathan used a diffuse reflection light source to obtain a thread profile projection, and proposed a profile processing algorithm to estimate the thread parameters, as in [12]; Chen et al integrated the photoelastic effect and an image processing algorithm to measure the contact angle of the ball screw, as in [13]; Li et al proposed a Res Unet-based thread edge recognition method that eliminates the need for thread area calibration and identifies the thread edge in a complex environment, as in [14]; Chen took the possible thread shape distortion during CCD shooting into account, and gave a corresponding compensation algorithm for the image distortion on the optical angle, as in [15]. However, the above solutions are only applied to the detection of ordinary threads with a linear profile, while the Journal homepage: https://content.sciendo.com thread profile of the ball screw is a curve, which is more difficult to detect.…”