A reference to compare against
01
The drawing, the model or an approved sample, defining what correct looks like. Tolerances have to be stated, because a comparison without a tolerance produces a difference rather than a verdict.

Advanced Manufacturing · Inspection
Automated visual inspection is the use of cameras, and in many systems lasers alongside them, to check manufactured items against a defined standard without a person carrying out the examination. The system captures the item, compares what it sees against a reference, and returns a result.
The reference is the part that decides whether it works. A system compares against something: a drawing, a design model, or an approved item held as a Golden Sample. Without a reference there is imagery but no inspection. Zabidou compares against a Golden Sample or the drawings and returns a pass or fail at the station, on every item that comes off the line.
Zabidou in production today
QLD + VIC
Operating in factories across Queensland and Victoria today.
100%
Inspection coverage on finished items as they come off the line.
Silver
Associate Member, Australian Steel Institute.
Bronze
Member, National Precast Association Australia.
Automated visual inspection is often described as a camera problem. In production it is three problems, and the camera is the one most easily solved.
01
The drawing, the model or an approved sample, defining what correct looks like. Tolerances have to be stated, because a comparison without a tolerance produces a difference rather than a verdict.
02
Plants have dust, vibration, variable light and temperature. A method that works in a laboratory and not above a production line is not an inspection system.
03
An answer fast enough and clear enough to change what happens next. A deviation map that needs interpreting is data; a pass or fail at the station is a decision.
The technology is mature and the claims around it are not always. Three limits are worth stating plainly, because they decide whether a system is useful on a particular line.
01
An automated system finds deviations from its reference. It does not notice that something is wrong in a way nobody anticipated, which is a thing experienced inspectors do routinely.
02
Optical methods measure geometry and surface condition. They do not establish material strength, internal condition or anything beneath the surface, and no amount of resolution changes that.
03
Where the output goes to a screen nobody watches, or to a report read weekly, automation has changed how the measurement is taken without changing when the decision is made.
The value of automating an inspection comes less from the accuracy of the measurement than from where in the process it sits, which is why Zabidou puts it inside the production step rather than after it.
A manual check is a separate activity. Someone stops, examines, records and moves on, and the frequency of that activity is set by the time available rather than by the number of items produced.
An automated check placed inside the production step runs because the item passed through, which means coverage stops being a decision. Every item is measured, and the measurement exists whether or not the line was busy that day.
Zabidou retrofits onto an existing line and reads every item as it comes off, returning a pass or fail at the station and keeping the record against that item. It uses lasers and cameras, and the comparison is made against a Golden Sample or the drawings.
For a plant the gain is that coverage stops being a function of how busy the shift was. Every item carries a measurement, a developing fault is visible on the second unit rather than the thirtieth, and the inspector gets the numbers in seconds while the item is still in front of them. The same check runs on a precast concrete bed and on a steel fabrication line.
Inline checking covers the same ground as a 3D laser scan at line speed, picks up defects such as honeycombing at demould, and puts first pass yield on a measured basis.
Five checks run inside the production step, on every item that passes it. Between them they cover the geometry of the item, the position of each feature and the condition of the surface, which is the ground optical inspection can genuinely cover.
Complete 3D measurement and verification of as-built output against a Golden Sample or design intent, for full dimensional compliance without stopping the line.
Automated, real-time verification that components are correctly positioned before and during production, while the mould is still open or the section is still in the jig.
Inline detection of surface defects, voids and irregularities at industrial speed, fully automated and without manual inspection.
Precise measurement and verification of critical dimensions against a Golden Sample or the drawings, inline, in real time, at scale.
Manufactured-unit centric digital provenance that automatically delivers the full audit history operations, financiers and customers demand.
Not a comparison of accuracy. A comparison of coverage, timing and what survives afterwards, which is what usually decides the outcome.
| Manual, item by item | Sampling a proportion | Automated, inline | |
|---|---|---|---|
| What gets checked | What there was time for. | A defined fraction of output. | Every item that comes off. |
| When the result exists | At the moment of checking. | After the sample is assessed. | As the item is produced. |
| What it depends on | Inspector availability and attention. | The sampling plan holding true. | The line running. |
| What is left afterwards | A signature. | A statistical statement about the batch. | A measurement against each item. |


It is the use of cameras, often with lasers, to check manufactured items against a defined reference without a person performing the examination. The system captures the item, compares it to a drawing, model or approved sample, and returns a result.
It is sometimes called machine vision inspection or automated optical inspection, and in electronics manufacturing the term AOI is standard. The principle is the same across all three.
The item is captured by one or more sensors as it passes a station, the captured geometry or image is aligned to a reference, and differences beyond the stated tolerance are flagged.
The accuracy of the sensor matters less than the quality of the reference and the tolerances, since those decide what counts as a deviation.
Anything that is not visible to an optical sensor. Internal condition, material strength and subsurface flaws are outside what cameras and lasers can establish.
It also only finds deviations from what it was told to look for, so an unanticipated problem may pass where an experienced inspector would have paused.
It replaces the measuring instrument rather than the judgement. The decision about a doubtful item, the disposition of a non-conformance and the reading of an unexpected result all stay with the inspector.
What changes is that the inspector works from measurements taken on every item rather than from the ones there was time to check.
Retrofit systems exist for exactly this reason. Zabidou fits onto a line already running rather than requiring the line to be rebuilt around the inspection step.
Whether a retrofit suits a particular line depends on the space at the station, the speed of the line and where in the sequence the measurement has to sit.
Ready to explore how ultra-vision can transform your operations? Send us a message and our team will be in touch.