OverviewAutomation of gear inspection has proven to be one of the most challenging in visual inspection. This is due to the individual differences in metal surfaces and the difficulty in collecting defective data. We have developed a gear imaging system that is compatible with the photometric stereo method and ImagePro, which significantly reduces the lead time for automating gear inspections.
When metallic reflections, surface patterns, etc. exist, the training data for AI to detect defective parts requires about 10,000
images for each defect, which is difficult to prepare. We would like to achieve AI by devising a way to take pictures.
By photographing gears using the “photometric stereo method,” the impact of shadows, weak illumination, and metal reflections are suppressed. As a result, the number of training images is significantly reduced, enabling the introduction of AI into the gear inspection process.
Reduces the number of training images by 99%
Line Scan Camera
Area Scan Camera
Arrows (green) indicate defective areas. The left column shows normal imaging, and the right column shows imaging using the photometric stereo method.