Seals and pouches
Open seals, product in the seam, or a channel in a pouch. See Food and the medical device pouch seal.
Vision inspection and measurement that catch bad packs before the case packer.
Fill, seal, cap, count, placement and damage, checked on every pack, with a reject timed to the pack and confirmed when it fires.
Problems
What this station catches.
Fill level
Packs and bottles under- or over-filled. See Food and Beverage.
Count and missing items
Short packs, cases, and kits. See Food, Packaging and Medical devices.
Label placement (360°)
Missing, skewed or wrinkled labels. See Beverage.
Containers
Cracked or dirty empties. See Beverage.
Product appearance
Shape, color, breakage, bloom, and enrobing. See Food.
How it works
How the station checks.
Single-camera station
Faults that show from one side: fill, seal, presence, count.
Three-camera station
Faults that can be on any side: wrapped labels, caps, cosmetic marks.
A timed, confirmed reject
The reject is timed to the part and confirmed when it fires.
A record of every reject
Every reject logged with image and lot. See Traceability.
Pass or fail
From picture to reject.
01
Capture
A photo eye or encoder triggers the camera, and a short strobe freezes the moving pack. Filters even out brightness and noise when the check needs it.
02
Locate
The software finds the pack, so the checks land in the right place on a pack that arrives turned or off center.
03
Measure
Pixels become numbers: fill height, piece count, seal width, the size of a mark.
04
Decide
Each number is compared with limits set from real good and bad samples. A fail fires the reject when that unit reaches it, and the image and reason are saved.
05
Rules or deep learning
Rule-based tools suit measurements, counts, and presence, and each reject traces to a limit. Deep learning suits faults that vary in look, such as cosmetic marks and wrinkles, and trains on many labeled pictures. Many stations use both. See How machine vision works.
Measurement
How does machine vision measurement work?
The camera turns edges into numbers, and each number is compared with limits set from real good and bad parts.
Edges
From pixels to millimeters
Edge tools find the part outline to a fraction of a pixel. A calibration target imaged at the same distance converts pixels to millimeters.
Optics
Telecentric or standard lens
A telecentric lens keeps the size constant when the part moves toward or away from the camera. A standard lens needs the part at a fixed distance.
Lighting
Backlight for outlines
A backlight turns the part into a sharp silhouette, the most repeatable way to measure outer dimensions. Front light is used for features on a surface.
Resolution
Pixels across the view
How fine the station can measure depends on the pixels across the field of view. The imaging calculator shows the trade-off for your part.
Dimensions
What can a vision system measure on the line?
Size
Length, width and diameter
Outer dimensions of parts, closures and packs, checked on every unit.
Heights
Cap, fill and seam height
Cap seating height, fill height and can seam height, seen from the side. See fill level and seam inspection.
Position
Placement against a reference
Label placement, print position and part location measured from a reference edge.
Gaps
Gaps, pitch and spacing
Gaps between parts, the pitch of repeated features and spacing in a tray or on a web.
Line speed
What does line speed change for measurement?
Motion
Freeze the part
A short exposure and strobed light keep edges sharp, because blur makes a part look bigger than it is.
Presentation
Tilt and rotation
A part that tilts or turns changes its measured size. Guides, fixturing or software correction keep the view consistent.
Tracking
Encoder-timed reject
An encoder tracks each part from the camera to the reject, so the right part is removed.
Interface
GigE Vision cameras
GigE Vision cameras feed the vision controller, which sends pass or fail to the PLC for every part.
Limits
What can’t machine vision measurement do?
Hidden features
Only what the camera sees
A 2D camera measures the visible outline. Internal dimensions and depth need a 3D sensor, another view or an off-line gauge.
Release checks
Not a lab gauge
Inline measurement checks every part against limits. It supports, and does not replace, the calibrated gauges and lab checks used for release.
Limits
Set from real samples
Limits are set on the HMI from good and bad samples. The study’s test images show how fine the station can measure your part.
FAQ
Questions about machine vision measurement
QHow accurate is machine vision measurement?
It depends on the field of view, the camera resolution, the lens, the light and how the part is presented. The feasibility study measures your sample parts and states in writing what the station can repeat on them.
QCan vision replace calipers or a lab gauge?
For checking visible dimensions on every part on the line, often yes. For internal dimensions or the measurements used to release product, the lab gauge stays. Vision measurement supports those checks and does not replace them.
QDoes the part have to stop to be measured?
Usually not. Strobed light freezes a moving part, and an encoder times the reject. Some fine measurements need the part held still or indexed, which the study tells you.
QCan it measure height, depth or flatness?
A 2D camera measures what shows in its view, such as a cap height seen from the side. Depth, flatness or a surface profile needs a 3D sensor, which the study recommends when the job needs it.
Every JM station is handed over with the same documents, test and training. See what is handed over at site acceptance.