STUDY OF AN IMAGE RECOGNITION SYSTEM: IDENTIFICATION OF PART THREAD QUALITY BY CHANGING THE ANGLE OF THE PART WITH THE BASE OR THE ILLUMINATION OF THE PART

Abstract

Industrial image processing is mostly based on the use of special cameras or imaging systems installed within the production line (SmartRay 2023). Image recognition is a computer vision program that uses machines to identify and classify specific objects, texts, and digital images and videos. Basically, it's the ability of computer software to "see" and interpret things in a visual medium the way a human can (GLOVER et al., 2023). To find out the influence of the lighting and/or viewing angle of the examined part on its quality, the image recognition camera used in the study was the "SICK Inspector PIM60" image recognition camera, the "YONGNUO" LED YN-160S” light source and the camera dome and filters of various colors: red, blue, green, transparent, and variable ND filter. The results obtained show that using isolated lighting and clear, red, blue, green, and ND2 - ND64 filters, the recognition accuracy reaches only 3 recognitions out of 10 attempts, which is still only 30 % accuracy. The recognition of a defects with settings at the angle (10, 20, 30, 40 degrees) using clear filter and the external lighting from one side of the part is high-quality when the surfaces of the parts are completely clean without any mechanical damage (scratches, knocks).Only 3 out of 10 parts threads features were correctly recognized, as the camera captures additional reflections from the part surface because of additional damage on the parts surface (Gasiūnas et al.,2022). The obtained results show that using a green camera filter and external illumination of the part from one side is the best way to identify the thread quality of the part. The results obtained show that using a green filter and external illumination of the test piece, it is possible to obtain ~ 100 % recognition of the part and its holes dimensions quality if the part is always placed in the same place and the same lighting conditions are used (Gasiūnas et al., 2022).