Vision inspection that catches the defect a tired eye misses on the thousandth part
What a vision system can check
| Presence and absence | Is every component fitted — clip, O-ring, spring, screw, label? |
|---|---|
| Orientation | Is it the right way round, the right side up, in the right position? |
| Codes and text | Read barcodes, Data Matrix and printed text, and check them against the expected part. |
| Dimensions | Measure diameters, gaps and positions within the camera's resolution. |
| Surface defects | Scratches, dents, cracks, rust and burrs — the hardest checks, decided by lighting. |
The record behind it
Our founder engineered automated visual inspection for two global bearing manufacturers, delivered through an automation firm: camera selection, lighting design, inspection logic and integration with Siemens S7 and Allen Bradley PLCs, through to operator training. Those systems reached 98% defect detection accuracy and cut manual inspection cost by 35%.
How a vision project should run
- Samples first: good parts and every kind of bad part, labelled.
- A lighting and camera trial on those samples, before any system is quoted.
- Agreed numbers: which defects, how small, how fast, and the acceptable false-reject rate.
- Integration: trigger, result and reject handling in the PLC, and images saved for every reject.
- Validation on production parts, and training for the operators who will live with it.
Good ones and bad ones. We will tell you whether a camera can catch the defect, and what the system would involve.
Machine vision questions
Can a camera check our part?
Send good parts and, more importantly, bad ones — every kind of defect you need to catch. A feasibility trial with real lighting on your real parts answers the question before anything is quoted for a full system.
Why does lighting matter so much?
A camera can only judge what the light shows it. The same scratch is invisible under diffuse light and obvious under a low-angle light; a hole is easy against a backlight and hard from the front. Most failed vision projects failed at the lighting, not the camera.
What happens to a rejected part?
The PLC acts on the result: a reject gate, a diverter, a stop with an alarm, or a mark on the part. The image of every reject can be saved, so a quality engineer can see why it failed.
Can vision be added to an existing machine?
Usually, if the part can be presented to the camera in a repeatable position and there is a moment in the cycle to take the picture. The trigger, result and reject handling are added to the existing PLC program.
Have a job that a machine should be doing?
Send the part and the target output. You get a concept, a sequence and a budget range back — free, and with no obligation to build it with us.