Abstract
Films are used in various products such as liquid crystal displays and flexible printed circuits. In the film production processes, the film is eventually wound to form a roll, but in the film winding process, defects such as wrinkles and scratches may occur. Although defects make the film roll worthless, resulting in a large manufacturing loss, conventional inspection methods cannot inspect them inside the film roll due to mostly the surface measurement. Therefore, we proposed a three-dimensional imaging inspection for film rolls using optical coherence tomography (OCT), which is a three-dimensional imaging modality with high spatial resolution mainly used in clinical diagnosis. Our previous study demonstrated that the occurrence of defects and types of defects can be inspected from tomographic images of film rolls visualized by OCT. In the case that OCT is employed in the film winding process in the production line, however, it is difficult to inspect the images visually since films are wound at high speed. To address this issue, this study proposes an automatic inspection for the occurrence of defects by using OCT tomographic images of film rolls and an unsupervised machine learning technique. In the presentation, we will show the results where the defect detection accuracy of the proposed method was investigated against the film thickness, the film material, and the defect type.
