Camera Module Surface Inspection
Ensure that phone camera lenses are free of surface defects
A breakthrough in complex inspection, part location, classification, and OCR
Before being installed, a camera lens must be clear of fibers, scratches, smearing, fingerprints, dust, or other foreign objects. It must be checked again after installation to ensure that no contamination or damage occurred during installation.
Scratches can appear anywhere on the lens, in a variety of orientations. Contaminants can vary widely in appearance and location. Reflections from the lens can confuse traditional machine vision while hiding various defects. Along with all these possible issues, the transparency of the lens itself means that refracted images from objects below can create a confusing background. Traditional machine vision cannot be programmed to identify the wide range of defects against the confusing and ambiguous background, often requiring additional manual inspection. Since surface defects can cover a wide range of unpredictable variations, deep learning inspections offer the most robust solution for inspection.
Cognex Deep Learning’s defect detection tool trains on a wide selection of both defect-free and defective lenses to learn the full variation of normal parts. As it scans through a sequence of lenses it analyzes and flags any that are outside of the acceptable range, while minimizing false positives.
If future lens models change, the defect detection tool can quickly train on a set of images to quickly begin inspecting the new lens with minimum delay. Cognex Deep Learning’s classification tool can be used after defect detection to distinguish between easily reworked defects such as fingerprints or dust and identify more significant defects that require corrective action.