Project brief
This proposed bench would explore visual inspection of simple gaskets placed in a generic housing groove. The study would combine a camera fixture, controlled illumination, and a local image-processing pipeline. Rather than promising universal defect recognition, it would start with a limited set of visible conditions such as missing segments, obvious displacement, and incomplete seating in a transparent demonstration fixture. Registration features would align each image to the same reference before region-based checks are applied. Proposed deliverables include the imaging layout, an annotation guide, and a review interface that preserves the source image. Performance would be evaluated only against a labelled bench dataset, and the concept would not constitute an automated production acceptance system.
Design concept developed for this portfolio. The diagrams describe the proposed system; implementation and testing are part of the validation plan.
The engineering challenge
Separate visible gasket-placement differences from changes caused by illumination, camera position, and part appearance.
Engineering approach
- Create a repeatable camera and lighting fixture.
- Define visible conditions and an annotation guide.
- Register images before applying region checks.
- Preserve uncertain cases for manual review.
Validation plan
- Use labelled samples representing each defined condition.
- Vary lighting and placement to inspect sensitivity.
- Report false positives and missed visible conditions separately.
Concept scope
- Controlled imaging fixture
- Reference registration
- Region-based checks
- Annotated sample review
- Uncertain-case queue
Software & engineering tools
OpenCV, Python, FreeCAD, LED illumination, USB camera