KOMPASS brings label placement inspection to every container leaving the labeller. AI vision confirms each label is present, sits at the right height and position and lines up with the container, then signals the line before misplaced labels reach the case packer.
Labellers apply thousands of labels an hour. When the applicator shifts, every container after that moment carries the same fault, and the shelf shows it straight away.
Our AI Product Inspection system, KOMPASS™, performs label placement inspection with models trained on your containers and labels, locating each label and measuring its position against the container. Positions are logged with an image of every container.
Jidoka's labelled-bottle demonstration shows label placement inspection in practice: custom AI models check the quality of label application, together with OCR and OCV reading of printed text, so every bottle is checked for how its label was applied and what it says.
Cameras after the labeller capture each container. Round containers may need several views or rotation to see every label.
The model finds each label and the container's own reference points, then measures height, offset and angle.
Failures are rejected, position trends flag drift and every result is stored with its image.
Send us your containers, label specifications and examples of misplaced labels. We will assess camera positions and lighting for your labeller.
Get In TouchEach method needs a clear view of the label and container. The difference is whether label placement inspection measures position and sees drift early.
Jidoka's labelled-bottle inspection for Diageo checks orientation, text and label condition in-process at 300 bottles per minute, with automated ejection of faulty bottles.
Diageo's published results include label orientation checks within its complete labelled-bottle inspection.
Label placement inspection measures where each label sits on the container: its height, horizontal offset and angle, along with whether it is present at all. The measurement is taken from the container's own edges or features, not from a fixed point in the image. That way, normal container movement on the conveyor does not cause false rejects, while a real shift in the label does. Each container gets its own measurement, so one badly placed label is caught even in a run of good ones.
Yes. Front and back labels, neck labels, collars and tamper strips can all be checked in the same pass, each with its own position limits. The station also confirms they are on the correct sides, which catches labels swapped between sides after a changeover. Round containers may need several cameras or a rotation step to see every label. Each label type has its own limits, because a neck collar and a front panel have different tolerances.
A single camera sees only part of a round container. Several cameras around the conveyor, or a 360-degree view created by rotating the container, show the whole label. Where containers are not oriented on the line, the model finds the label wherever it is and measures its position relative to the container, instead of expecting it at one angle. Where every label must be seen in full, rotation gives the most complete view.
Common causes include changes in web tension, worn belts or rollers, a shifted label sensor, applicator pad alignment and adhesive temperature. Container variation and line speed changes play a part too. Because these causes act gradually, applicator drift shows up as a slow trend in measured position long before labels fall outside the limit. Seeing the trend early means a quick adjustment instead of a stop to relabel or rework product.
Yes. The station reports label height and offset as numbers, so operators can see which direction labels are moving and by how much. Feeding that trend back to the label applicator settings corrects the cause instead of rejecting containers one by one. Where the labeller accepts it, the correction can be automated. Operators can see the effect of each adjustment straight away in the next readings.
A sensor for label presence confirms something is there. Missing label detection with vision also confirms it is the label, in the right place, and not a partly applied or folded label that a sensor would pass. The same image is used for position checks, so one station covers presence and placement without extra hardware on the line.
Start from your label specification and the point where a misplaced label becomes visible or hides required information. Height limits for neck collars and offset limits for front panels are often different. Validate the station with containers just inside and just outside each limit, and adjust lighting if the label edge is hard to see against the container. Limits can be tightened later once the labeller is running consistently.
Send examples of your containers and labels, the position specification and photos of typical placement faults. Add the line speed, the labeller layout and where a reject can happen. With that, we can recommend camera positions, lighting and the container reference points for label position inspection on your line. Photos of past placement faults help us set limits that match your quality standard.
Share your containers, label specification, labeller layout and line speed. We will plan label placement inspection after your labeller and walk through cost and schedule. Expect about 6 to 8 weeks to go live.
