A system for detecting scratches on the surface of small-sized metal shafts after the grinding process using optical techniques combined with computer vision.

Abstract During the grinding process of metal machine parts requiring a high degree of smoothness, the use of diamond-surfaced grinding tools, specialized grinding stones, or high-hardness metals are common choices. However, during grinding, surface tool damage or misalignment between grinding surfaces can cause defects such as cracks, scratches, or complete surface destruction. Detecting small scratches is one of the biggest challenges due to their size and distribution, making it difficult to quickly identify them using the human eye. Automated systems, on the other hand, struggle to highlight scratches because automatic optical lighting systems face strong reflections from the metal surface, especially cylindrical surfaces after being polished. As a result, the images obtained lack controlled reflection positions and contrast between normal and scratched areas, leading to inefficiency in detecting these scratches. In this study, we propose an optical system combined with lighting techniques to highlight scratches on the surface of a cylindrical machine part after polishing, thereby detecting and classifying these defects to make pass-fail decisions in practical production line applications.