Computer vision application
Counting and measurement with computer vision
Counting parts and measuring dimensions by camera is a different problem from detecting defects: the criterion is numeric, the error is measurable, and calibration matters more.
Tell us what needs counting or measuring, and to what tolerance.
What can be counted
Automatic counting by camera, with a record of the event behind every number.
-
Parts on the conveyor
Continuous counting of moving items, logged by shift, batch or product.
-
Items per package
Quantity verification before sealing, preventing incomplete boxes.
-
Stage inputs and outputs
Volume entering and leaving each segment, to locate where loss happens.
-
Events, not only objects
Cycles, actuations and recurring occurrences that nobody records today.
What can be measured
Measurement from a calibrated image, with no contact with the part.
-
External dimensions
Length, width, height and diameter from a calibrated image.
-
Area and fill
Coverage, fill level and the proportion occupied relative to the standard.
-
Distance and alignment
Gap, spacing and positional deviation between components.
-
Occurrences per part
How many defects appear on a single unit, not just whether one is present.
What determines accuracy
The useful question isn't how accurate the system is, but which error your process accepts.
-
Calibration
A pixel measurement only becomes millimeters with a known reference and calibration maintained over time.
-
Distance and optics
Variation in the distance between part and camera changes the scale. Lens, focus and perspective all enter the error budget.
-
Part stability
An object that wobbles, rotates or overlaps another calls for synchronized capture or more than one camera.
-
A stated tolerance
We start from the tolerance your process accepts and check whether the arrangement can reach it — not the other way around.
The number as a record
Counting and measurement only become a trustworthy indicator when they stay tied to evidence.
-
A number with evidence
Every count or measurement can be stored alongside the image that produced it.
-
Auditable history
Date, time, line, product and classification stay available for later verification.
-
Plant integration
Values feed MES, ERP or SCADA, or a dedicated dashboard when no destination exists.
Frequently asked questions about counting and measurement
What production and quality teams usually ask.
What measurement accuracy is achievable?
It depends on calibration, optics, distance and part stability. We work backwards: we start from the tolerance your process accepts and check whether the arrangement can reach it.
Does it work with overlapping parts?
It's harder. Overlap usually calls for a different angle, more than one camera, or moving the point on the line where counting happens.
Can it count and inspect at the same time?
Yes, and it's common. The same capture can feed counting and defect detection, as long as the image meets the requirements of both.
Does it replace a scale or sensor?
Not always. When the quantity is weight or something not visible, the sensor remains the right tool. Computer vision covers what is observable.
Talk to our team
Let's understand your challenge
Prefer to book the diagnosis conversation? Use the scheduling link. Or fill in the form — our technical team replies.