Application of Methods for Detecting the Boundaries of an Object in an Image to Search for a Pilot Line

The article considers the possibility of using computer vision for the task of detecting a pilot line on a pipe. The pilot line is applied with paint during the production process to the surface of the pipe being welded from the sheet. The line is rotated at an angle of 90 degrees (at the three o'clock position) relative to the weld and parallel to it. Thanks to it, it is possible to position the installation of local heat treatment of the weld, due to which residual stresses formed during welding are removed. The method of image classification based on the weighted average brightness of the image is considered, three methods of detecting the boundaries of an object in the image are analyzed (the threshold binarization method, the Canny Edge Detector edge detection operator and the Gabor filter), as well as an algorithm for selecting straight lines using Line Segment Detector. An assessment of the quality of work of each of the methods in solving the task was carried out. The best result was shown by the edge detection operator Canny Edge Detector. A Python program was developed using the OpenCV library, the Kivy framework and the KivyMD add-on.