Check carton label placement by finding a stable reference on the package and comparing the visible label edges with an agreed tolerance window. Mark closely borderline and obscured examples for review. Evaluate the check on approved formats and ordinary variations in how the carton sits.

Set the reference before measuring position

Use the package face and its corners when the carton can move inside the camera frame. Suppose a fictional operator lines up labels using the front face’s lower-left corner. A camera that compares the label to a fixed pixel point may produce different results when the carton shifts away from the fixture. Reference the agreed corner or constrain the physical positioning so the camera view and product geometry have a repeatable relationship.

Define horizontal displacement, vertical displacement, and rotation if the requirement covers them. A label can be centered and still touch the window at a rotated corner. Keep placement and print checks distinct. Text style, legibility, and any barcode measurements have their own rules and evidence. A position result does not establish these properties.

Decide the boundary with real examples

Imagine an approved fictional tolerance of 2 mm horizontally and 3 mm vertically for visible offsets measured relative to the baseline position. An offset of −2 mm horizontally and +2.4 mm vertically is within those limits if the boundaries are inclusive. An offset of +2.2 mm and +1 mm is outside the horizontal limit. If an edge is hidden, the worksheet cannot establish the missing measurement.

Fictional sampleMeasurementsAction
Matte carton, slight shift−2.0 mm horizontal and +2.4 mm verticalWithin the fictional inclusive window
Matte carton, offset label+2.2 mm horizontal and +1.0 mm verticalCompare the visible edge against the horizontal limit
Glossy carton, reflected lightLower label edge unreadableAsk for a second view or human measurement
Fictional offsets, based on a ±2 mm horizontal and ±3 mm vertical placement tolerance.

Check how the operator measures a real boundary case. If measurement resolution or positioning introduces variation close to an edge, specify what ambiguity goes to review. Agree on whether the label should be measured in the camera plane or normalized to the package face. Use appropriate references and measurements for your own line. The label position worksheet compares supplied offsets but does not measure a photo.

Test enough ordinary variation to learn where the rule applies

Capture accepted and off-center labels on every included carton size. Test the lighting and fixture across ordinary shifts. Add examples near the tolerance and contrast them with acceptable tilted packaging. Preserve the original and review measurements. If one material makes a corner hard to see, group those samples and report the need for alternate evidence.

Look for patterned mistakes in the pilot report. False flags that appear across allowed carton tilt may indicate a geometry problem. Missed defects on glossy stock could indicate a hidden edge. A persistent disagreement at the tolerance could reveal that the reference method is inconsistent. Make distinct follow-up decisions for each observation rather than changing every threshold in response.

Keep the check attached to the package and next action

Record which SKU and label format each evaluated window covers. Give reviewers a named alternative when a carton has left the approved view. For example, an operator might keep the original capture, select a second view from the same unit, and attach the final decision to that unit. Test whether this works at normal pace and where rejected or unclear items are expected to go.

Compare the controlled example in the fictional inspection walkthrough with the requirements on your actual line. Cognex’s image acquisition and part placement documentation offers a platform example of keeping the workpiece positioned for reliable capture.

Turn the drawing into a measurement plan

Start with one carton, the placement drawing that governs it, and a representative finished sample. Put them side by side with the person who accepts or holds the carton. Ask them to point out exactly where the dimensions begin and end. The drawing may locate the bottom and left edges, while the shop floor fixture references a crease that occasionally folds out of view. A camera recipe that follows the fixture when the quality decision follows the crease can pass the wrong item with a perfectly consistent reading. Put the measurement reference in the scope of the check before choosing a camera, light, or inspection method.

For each included label format, write down the approved package face, the reference edges or landmarks, the label features to find, and the expected physical location. Include which side the viewer is looking at and how the test reports sign. A positive horizontal displacement might mean that the label moved toward the right edge in the operator’s view. A downstream dashboard may use a different coordinate convention. Preserve the convention next to the results so that its receiving team can interpret a signed pair like +1.6 mm and −0.3 mm and can trace it to the photo.

Some drawings use the center of the label relative to a dimensioned point. Others protect an entire rectangle or a physical feature, such as a tear strip. A center estimate can discard important evidence about rotation, shape, and corners. If the requirement is that the label stay within a window, identify the parts of the outline that establish that result. A corner that crosses the protected edge matters even if the center is precisely at the desired point. If the drawing separately specifies centering and any protected overlap, establish each result and define the final action when their outcomes differ.

Reserve a line for what the check is allowed to conclude. Typical components of a scope record are a carton family, a view, a material or finish, a station, a label format, a measurement rule, and a decision owner. A finding on one matte carton supports that configuration. Demonstrate any intended extension across size, material, camera setup, or label artwork before applying it. Keep print readability and barcode grading in their own scopes if the buyer needs those checks. The same station can support several decisions only if its evidence and actions are clear for each.

Work in carton coordinates when the carton can move

Consider a fictional setup in which a flat carton face fills the center of the frame. When the carton moves sideways by ten pixels, its reference corner and the label both move by ten pixels. A fixed image-space box sees a different position. A measurement referenced to the carton should stay stable if the rest of the setup is unchanged. This simple paired comparison is a useful diagnostic before spending time building a more complex recognition method. Capture the same item in several allowed positions and ask which aspects of the image move together.

Rotation needs more care. Move the physical carton within its usual guide and look at a pair of reference directions. A line across the lower carton edge and a line across its left edge may rotate together around the camera view. The label stays in place on the carton when both sets of edges rotate together. An inspection can estimate the carton orientation and measure relative rotation after finding stable geometry. If the box corners are hard to locate when damaged, hidden, or behind a guide, describe the fallback that reestablishes the measurement or sends the unit to review.

Perspective can change the shape of a rectangular label in a photograph. A carton leaning forward might make one side appear shorter even if the real label is unchanged. A stable arrangement can limit that movement. When the station has a legitimate range of pose, compare a mapping from the observed surface to the carton measurement plane, using known geometry and representative physical references. Check the resulting dimensions against a reference measure for the positions that actually occur. Ask how the method behaves on a warped flap or rolled edge where the planar assumption breaks.

Keep the transformation visible in the evidence. Save the original image and the coordinates used for the reported measurement, along with the check revision. Show the identified carton landmarks and label landmarks as a diagnostic overlay. In a disputed result, the original can reveal that the check chose a printed line or the edge of a tape seam in place of the intended feature. The computed displacement alone cannot resolve that issue. A reproducible overlay lets the setup owner decide whether the rule, the image, or the underlying feature needs work.

There is a sensible place for a fixed region. A fixture that pins every unit against the same pair of physical stops can make the window easier to locate. Prove that units reach the stops when running normally, and test one known placement in positions near the limits of allowed presentation. A fixed crop might also serve as an early gate to confirm that the expected carton face is in view before a measurement takes place. Record the assumptions so a later fixture adjustment does not silently change the recipe.

Compare edge choices under ordinary variation

Prepare a paired sample using a carton with a strong label outline and a carton where the print nearly matches the substrate. Photograph each in its centered position and at the allowed edges of its station movement. Look for locations where the outline, carton edge, and print artwork compete. A diagonal artwork feature is an easy false reference if the candidate search extends across the entire label. A material transition, a printed border, and the applied label edge can all appear near the same expected location. Have the quality owner identify the intended physical line before interpreting the overlay.

A good test set also includes variation within a known finish. Matte board can differ in tone, stiffness, or fiber pattern. Two batches of the nominally same label can differ in adhesive behavior or corner lift. If the rule follows the manufacturer’s cut edge, a small print registration change should have little effect while the label still meets the placement dimensions. If the customer’s actual requirement references printed artwork, document that basis instead and verify it against the drawing. Similar photographs can therefore produce different legitimate decisions across contracts or product groups.

Plan deliberate look-alikes. Pair a well-positioned label with a colored background stripe close to the edge and one with a real position error on an otherwise plain surface. Pair a label with a shadow under a lifted corner and another on a slightly curved panel with no lift. Retain the original and reference measurement, including how the second reviewer placed a ruler, gauge, or template. If two reviewers repeatedly choose different edges, resolve that issue before using the samples as the expected answers.

Add a view check to the setup. A readable stationary image can become too blurred at the intended operating pace if motion during exposure obscures a small edge. A strobe or changed exposure may improve the image while creating a new glare pattern on glossy stock. Work through lighting controls and assess the paired change on the same carton with stable ground truth. Hold product variation constant where possible so the impact of the setting change is visible in the comparison.

Set up a repeatability study around the actual limit

For a fictional horizontal limit of ±2 mm, take a small group of labels around the positive and negative limits plus clear in-window and far outside controls. Give each carton a unique ID and use references the evaluation team can trace. Repeatedly present the same cartons under expected position and orientation variation. Record the exact measurement and disposition every time. Interleave the cases so a station operator cannot infer the correct decision from the order of presentation.

Separate three questions in the notes. The first is whether there is a stable reference measure on the physical carton. The second is whether the camera measurement agrees within the intended physical range. The third is whether repeated presentations produce consistent actions. If human measurements disagree at the boundary, strengthen the physical reference, record an uncertainty band, or route those items to a second method. If image readings disagree while physical readings are stable, group results by pose, finish, rate, and the geometry actually found.

Imagine a reference reads +1.9 mm on a matte unit. Its camera readings over four presentations are +1.7, +1.8, +2.1, and +1.9 mm. The intervals overlap the +2 mm limit. A software change that rounds every reading down would conceal the crossing while undermining the original boundary. Detect which inputs produced the spread. A useful comparison checks whether the edge coordinates move because of blur, an angled face, shifting reflections, or changes in the detected landmark. Keep both the physical reference and the original photographs attached to the case.

A second fictional unit measures +3.4 mm and is flagged on every repeat. That is valuable as a protected outside case, but it says little about discrimination at +2.0 mm. Likewise a centered carton that is always clear tells you setup and evidence are working for that sample. Use the deliberate boundary group to learn whether measurements near the operational limit are consistent enough for an automatic action and how often those cases occur when the line runs normally. Keep boundary probes and ordinary prevalence samples identifiable in the results.

Pick a resolution that the setup can support. An internal result can keep more digits for debugging. Displaying three decimal places to an operator can misstate the certainty if their reference and camera repeatability are coarser. Agree on the rounding policy and whether the operational decision compares a rounded display number or an unrounded internal value. When a public or partner report shows a numeric result, accompany it with the units, limit, sign, and enough detail to explain an apparent mismatch between the display and the recorded decision.

Handle multiple axes and rotation explicitly

A simple centered placement example checks independent horizontal and vertical offsets. A different requirement can constrain the entire label within a marked zone, or cap angle separately. Translate the geometry into actual cases before choosing the calculation. Ask the quality owner what happens if both offsets are small but the angle crosses the tolerance. Ask about offsets exactly on the limits, opposing signs, and cartons where the top edge is visibly bowed. The answers should agree with the written requirement and the reference measurement procedure.

For a fictional example, a 60 mm wide label rotates a few degrees about its center. One lower corner moves down and its opposite moves up. If clearance above a tear strip is small, the rotated corner can cross the minimum clearance while the center moves very little. Evaluate the relevant corner relative to the strip, on the same panel surface. Keep the angle measurement for diagnosis if the requirement is clearance. A separate angle limit may still apply if the placement drawing specifies one independently.

Decide how checking order affects action. If the fixture check fails, a reported label offset can be undefined. If the face is fully visible and the label is physically missing, there is no label edge to measure. A missing label may justify a separate defect class when in scope. If the label might be hidden by glare or another part of the carton, treat that as an evidence problem and follow the documented review or recapture path. The worksheet needs a way to represent each of those cases without storing a zero that will later be interpreted as perfect placement.

In the station display, give each measurable predicate its physical reference and result. An overlay that simply colors the entire product red supplies little guidance about how to fix a shifted label. An operator can use a reason such as “left label edge outside the horizontal window” with a line showing the discovered reference and the failing boundary. A reviewer handling an unreadable sample can use “bottom edge hidden by reflection,” a picture of the relevant crop, and the route to recapture.

Exercise the check through a format changeover

Suppose the line switches from a shallow box with a label near its bottom edge to a taller one whose label sits above a shipping band. The first carton may fit the new camera crop even though the wrong placement window is active. Measure the startup cases against the true recipe identity and keep a record of the applied configuration. A confident software result from the wrong recipe is invalid for both the intended geometry and the evaluation denominator. Prevent a result from flowing into a normal action when the product format is unresolved.

Run a fictional changeover rehearsal with three kinds of samples. Use approved first pieces from each format with clearly different intended label locations. Use one item deliberately close to a boundary and one item that meets the previous format’s geometry but fails the new format’s requirement. Put one sample into the queue just before the change and check whether its outstanding decision retains the old recipe. Make the operators and the receiving system use only the information they would have in a normal changeover.

On startup, check that the right datum remains visible after adjustments to guide rails, camera aim, focus, lighting, or the position of the trigger. Verify whether the resulting crop changed and which evidence becomes unavailable. A replay of stored photos can test decision code but cannot prove that the new physical presentation produces the expected images. Collect current setup evidence as well as the historical reference set when both the environment and the software might have changed.

When the switch works, record who can approve restart, which first-piece cases were measured, where their reference decisions live, and which setup versions were active. Include a route for production when the recipe remains uncertain, such as conducting the approved manual check or holding the affected group. Coordinate with the team that controls disposition since the scope of a camera check does not itself authorize product movement or release.

Build a pilot set a reviewer can diagnose

Use paired controls: the same carton in several allowed positions and several cartons in the same position. The first family isolates capture variation. The second provides variation in labels, finishes, and physical samples. Add authentic accepted and rejected evidence where available, and label intentionally staged defect samples so they do not imply normal defect prevalence. Do not treat five images of one deliberately displaced carton as five independent misplaced labels. Keep related images traceable to the unit and collect independent evidence for the conditions you plan to approve.

Divide the examples into setup, tuning, and final evaluation groups as applicable. Introduce a deliberate protected sample for each check decision, including cases where a visible placement error must be flagged and a known bad view must request intervention. Put ordinary production samples alongside those examples to estimate how many units might be sent to review. Build the final set before inspecting outcomes and preserve the associated dataset split. If you make a new adjustment after reviewing a final group, document its role and collect independent confirmation of the updated recipe.

Choose condition labels a team can act on. “Glossy stock, low angle” can be compared against the same unit with a changed light position. “Unusual failure” may tell a reader little about what to reproduce. Useful tags for this check might include carton family, label finish, guide position, presentation orientation, view, actual line state, and setup revision. A small number of reliable labels usually supports more analysis than free-text opinions about every photo.

Describe missing or unavailable cases in the final packet. A known obstruction may be common because staff open the guide during changeover. A rare label rotation may have no independently measured production examples in the initial set. A high overall success rate can hide both conditions if you choose the denominator after the experiment. Use a transparent unit count for each evaluated class, preserve the staged reference group separately, and give the pilot owner a clear account of which scope still needs evidence.

Diagnose the patterns in a placement report

Consider a fictional pilot with three failures. The first group shows accepted matte cartons incorrectly flagged when they enter against the right rail. The detected label corners look correct, but the reference line now follows a line of print. The second group contains two off-center glossy examples whose bottom edges disappear under light. The check returned accepted after measuring a stronger nearby edge. The third group wants repeated measurements near the same limit, and its reviewers disagree about where an angled corner intersects the panel. Each incident calls for a different investigation.

For the matte rail cases, replay paired views of the same plus several independent cartons at the affected positions. Keep the acceptance rule and label measurements fixed. Compare candidate landmarks and restrict detection to the true reference features where appropriate. Re-evaluate clear controls so the change has not broken an allowed orientation. For the glossy cases, compare the crop, raw view, glare mask, exposure, or appropriate alternate view. A measurement returned with an unreadable required edge should be excluded from automatic acceptance and visible as an evidence failure.

For the boundary dispute, meet with the requirement owner and the original physical examples. Ask for a traceable method that two people can reproduce, including whether and how to straighten the carton. Use known units across both sides of the limit with enough variation to tell a decision rule problem from a measurement variation problem. A temporary review band can keep production decisions safe while you collect this evidence. Report how its use changes both operator load and the number of automatically clear or flagged observations.

Avoid collapsing all errors into an average distance. A handful of high deviations concentrated on a single product variant can reveal an unstable recipe even if mean distance looks small. A signed offset can also cancel across opposite directions. Show errors by direction and relevant condition, and separate wrong decisions from errors in readings that still landed on the approved side. Preserve unit IDs and group denominators so the next test can confirm whether the repair reaches the actual cases that motivated it.

Hand off the rule and the evidence for maintenance

Walk through a normal result, a true placement failure, and an evidence failure with the person who will use the station. Ask them to locate each carton, explain its status, and retrieve the picture without consulting the setup engineer. For the evidence failure, check that the documented secondary action is possible with the station’s actual space, time, and staffing. For the true placement failure, show what gets communicated to the process owner. A reproducible reason can help maintenance check guide movement or applicator alignment independently of the inspection software.

Record which values are editable and which owner has authority to change them. A recipe adjustment to improve edge finding does not imply the placement tolerance changed. A different tolerance belongs to the quality decision owner and may alter the accepted population. If someone enters an updated number on the station, make the version visible in the resulting records and keep the earlier version available to interpret previously held items.

Review the first production window under explicit scope. Compare accepted, flagged and reviewed samples with the agreed reference outcomes. Pay particular attention to first and last pieces after a changeover, pending reviews whose evidence was collected before the change, and the condition that created the most failed views during the pilot. Keep an estimate of the time spent retrieving units and making decisions. Give one owner the ability to suspend the automatic action when the required setup or evidence falls outside the approved range.

Prepare a one-page transfer record. It can contain the carton and label identifiers, physical placement drawing, reference geometry, known variations, visibility prerequisites, tolerance and rounding rules, expected outputs, reviewed examples, approved software and setup versions, and a recheck trigger. The examples give the receiving owner something to run again after replacing a light, moving a guide, or introducing a supplier finish. Put the underlying reference images and policy together so the procedure survives staff changes.

Practice with four decision records

Use any sample numbers here as fictional exercise inputs. For each example, write down whether you are judging a label, verifying sufficient evidence, or confirming that the recipe can run. Give a plain-language operator action and a named next test. Compare your answer with the explanations and replace the values with your own approved drawing and process rules when creating a production record.

A shift that follows the carton

Unit C41 has a left-edge physical measurement of +0.5 mm relative to the approved label position. It is captured at a centered rail setting and against the right guide. In the two frames its physical label edge moves by 16 pixels. The carton reference moves 16 pixels in the same direction. In the first frame the measurement overlay references the carton corner. In the second frame the overlay stops at a stationary fixture mark and reports an outside reading. The next step is to repair or bound how the reference geometry is chosen, retain both views, and challenge the repaired rule with independent cartons at both rail positions and true offsets. Use the real product decision only after the evidence is anchored to the correct datum.

A center inside and a corner across

Unit C72 has a center lying inside the approved placement window. Its label rotates clockwise. The lower-right corner comes closer than the approved clearance to a tear strip. The reference sheet names the minimum open distance to the strip as a separate requirement. An independent physical check confirms the encroachment. Record the center and clearance outcomes individually. Have the operator follow the approved nonconforming routine for the failed requirement. Include a similarly rotated label far enough above the strip as a contrasting control so the next release can show that the specific clearance is being measured.

A hidden measurement at normal pace

Unit C93 has a bright patch near its top-left edge when the line runs at usual pace. The camera can find the lower edge and right corner. The approved position rule still needs the obscured edge or a validated alternate where it is visible. The reading below the threshold came from an unrelated border in the patch. Route the unit for recapture or the manual method according to the evidence rule. Ask the station lead to keep the original and test different lighting with the same item before changing any geometry or numerical limit. Keep the example as a protected evidence gate if the alternate arrangement is approved.

A reused recipe at the wrong height

Unit C104 has the usual matte label on a box with a taller front panel. The view is clean and the check reported a number inside its current window. The system log shows the correct unit ID but a window belonging to the shorter box. The previous format’s check was active for several units after changeover. Preserve the original report and determine the affected time or product range from records you can verify. Use a confirmed format and an authorized review path for those units. Rehearse the changeover again under expected pace, preserve its approved first-piece examples, and require setup identity evidence before reporting the new result.

Review the findings before expanding

A practical placement check must give the same physical requirement the same treatment across the approved variation in presentation. An evaluation record should make clear how the datum was found, which edges were used, what range of repeatability was observed around the threshold, and how unresolved units were handled. Include what you learned from the cases the system got wrong and which underlying receipt or picture supports each conclusion.

Bring a first-piece checklist, the measured protected samples, and the open evidence gaps into your agreement about the next step. If reliable automatic decisions are established for only one finish or format, use that scope for the next controlled window. A material or layout change belongs in a new comparison whenever it changes what the station sees. Update the reference groups so operators and QA can recover the same intended checks from a later run.

Before closing the review, ask the quality owner to verify the acceptance window on paper and the station owner to replay one difficult unit through every affected action. Capture the active recipe identifier, the final operator action, and the applicable comparison set. Then test whether another shift can recover all three from the stored record. That handoff makes a visual placement result useful after the pilot team leaves.