Continuous surface inspection connects images of moving material to a decision about a specific strip, sheet, roll, or coil. Start by defining defects and coverage, then validate the imaging and decision system at representative production conditions.
This guide helps quality engineers plan inspection of continuous surfaces. Gather the material and finish range, width, line-speed profile, defect examples, acceptance rules, and the way affected material will be located and handled. The inspection system must preserve that identity from image capture through disposition.
Build a defect catalog with boundary examples
Define each defect class in terms of what matters to the product: type, extent, location, severity, and any customer-specific acceptance rule. Include acceptable texture, harmless marks, and borderline cases. Have qualified reviewers agree on the reference labels and record how disagreements are resolved.
A scratch, stain, pinhole, and local shape change may require different imaging conditions. The relevant question is whether the complete station makes the feature distinguishable across the expected material variation. A nominal pixel count or a best-case example is only part of that evidence.
On a narrow screen, scroll the table horizontally. Keyboard users can focus it and use the arrow keys.
| Dimension | Include | What it tests |
|---|---|---|
| Material and finish | Expected grades, colors, coatings, gloss, and normal texture. | Whether acceptable variation is confused with defects. |
| Defect class | Representative severity, size, orientation, position, and boundary cases. | Coverage of the actual acceptance rules. |
| Motion and position | Normal speeds, accelerations, edge positions, flutter, and changes in working distance. | Image geometry, blur, focus, and synchronization. |
| Operating condition | Relevant lighting variation, optical contamination, and production duration. | Whether performance persists outside a clean demonstration. |
| Material identity | Roll or coil changes, splices, restarts, and lost acquisition. | Whether a finding remains attached to the correct material. |
Make the defect visible before choosing the algorithm
Lighting direction changes what the camera sees. A geometry that emphasizes scattered light from a scratch may suppress other surface information. Brightfield, darkfield, diffuse, and backlighting arrangements answer different questions. Edmund Optics' illumination-geometry explanation demonstrates why lighting is part of the inspection method.
Evaluate representative samples with the lens, working distance, and mechanical arrangement intended for production. Record illumination settings, exposure, gain, and focus. If the required defect is indistinguishable in the captured image, collecting more algorithm-training labels will not supply the missing optical evidence.
The machine vision systems guide covers the broader imaging and validation workflow. For continuous material, add the motion and location requirements below.
Connect line rate to material speed
A line-scan camera builds an image from successive lines as material moves. At steady speed, the acquisition timing can be matched to that motion; with variable speed, encoder-based triggering can maintain the intended spatial relationship. Basler's line-scan use cases describe these configurations.
Check the camera's supported rate under the selected exposure, region, pixel format, and interface configuration. Basler's resulting line-rate documentation explains that actual acquisition settings can limit throughput. Also verify encoder scaling, trigger integrity, dropped-line detection, and downstream processing capacity.
Report misses and false rejects separately
Keep evaluation data separate from the material used to tune thresholds or train a model. Organize the evaluation by defect class and production condition. Similar adjacent images from the same defect should not create the impression of many independent test cases.
Preserve the location and the disposition
Join each finding to roll or coil identity, cross-web position, travel position, timestamp, image, defect class, and decision-version record. Define the coordinate origin and what happens at a splice or restart. Verify the physical relationship between the inspection point and any marker, cutter, reject station, or downstream review location.
Test loss of images, triggers, encoder signals, and processing capacity within the approved validation plan. The system should expose a coverage gap and route affected material according to the site's quality procedure. A quiet dashboard during an acquisition failure must not be interpreted as clean material.
Revalidate affected conditions after a lighting, lens, camera, line-speed, material, model, or threshold change. Retain a controlled reference set and representative production evidence. Plant-floor material visibility explains how findings and holds remain connected to the correct inventory.
Use a coverage and validation record
Download the surface-inspection validation record. It includes the line-rate and error-rate examples, defect-catalog fields, configuration identity, coverage interruptions, and the disposition checks needed before acceptance.