What should be inspected?
Define the inspection surface and the area where film should have been removed. Normal patterns, reflections and fixtures must be distinguished from actual residual film.
- Visible residual film or incomplete peeling
- Defect area and distribution within a defined ROI
- Good, defective and borderline sample comparison
From imaging to a decision
Film contrast varies with transparency, reflectivity and illumination. Trial imaging establishes whether the target can be distinguished consistently before selecting segmentation, area analysis or a learning-based method.
- Exclude fixtures and irrelevant background
- Evaluate defect area, shape and minimum size
- Validate false rejects and missed defects on independent samples
Residual film is not the same as adhesive residue
Adhesive, particles and film may look similar. Samples with confirmed defect causes are needed to determine whether the available imaging setup can distinguish them. A photograph alone does not establish material identity.
Preparing for line integration
Discuss triggering, stable acquisition, result handshakes and image retention alongside the algorithm. Minimum detectable defect size and cycle time depend on actual samples and imaging conditions.
- Provide good, defective and borderline samples
- Specify field of view, smallest defect and cycle time
- Confirm PLC interfaces and traceability requirements
Frequently asked questions
Is AI always required?
No. Stable contrast may support rule-based AOI. More variable textures can be evaluated for AI using representative training and independent validation data.
Can defect area be used for OK/NG?
It can be evaluated as a decision metric once the valid ROI, segmentation and acceptance limits are defined.