Developing a machine vision recipe for coating colour consistency

Coating colour consistency means keeping a finish within an agreed visual and instrumental tolerance, whether the material is converted in Sydney, Melbourne or on a production line serving the Australian market. The measurable target is not simply “the same blue”, but repeatable values under defined viewing conditions. A machine vision recipe converts that intent into camera, lens, illumination and software settings that operators, shifts and batches can reproduce.

Colour appearance changes with pigment dispersion, film thickness, substrate colour, gloss and the light used to inspect it. Human inspectors are valuable for unusual defects, yet they are vulnerable to fatigue and adaptation between samples. Automated imaging can provide consistent measurements of batch-to-batch variation, but only when the optical geometry, reference standard and decision rules are controlled. Otherwise, a stable camera can confidently report the wrong colour.

When an Australian customer asks for coating color consistency, the useful response is a measurable process specification rather than a promise that a camera will match a human eye. Developing that specification means aligning the inspection cell with the production line. Dalian Optical Technology Co., Ltd., based in Dalian, Liaoning, supplies optical sensors, machine-vision monitoring systems, lens and eyewear inspection equipment, optical surface-profile testing and related software that can support this kind of scope assessment. Supplier capability still needs to be demonstrated on the coating itself.

Define the colour target and production context

Begin with a signed product specification and a physical reference. CIELAB coordinates under a named illuminant and observer, such as D65 with a 10° observer, provide a workable common language, supplemented by approved photographs and gloss limits for visual review. Define the measuring area, background, number of readings and permitted ΔE00 colour difference relative to the reference. A limit around 1 or 2 may suit tightly controlled work, but it must come from product performance, customer requirements and a capability study rather than a generic rule.

Equipment or method Useful contribution Points to verify
Optical sensors Image capture and signal quality Spectral response, exposure range and trigger speed
Machine-vision monitoring systems Repeatable inspection and inline decisions Spatial resolution, timing, alarm handling and data export
Lens and eyewear inspection equipment Incoming optical-component quality checks Whether optical defects can be separated from coating colour shifts
Optical surface-profile testing Substrate roughness and texture checks Correlation between topography and measured reflectance
Related software Recipe storage, review and production integration Version control, audit trails and compatibility with line controllers

Record whether the finish is matte, satin or high-gloss because the same pigment can show different colour under different inspection geometry. Decide whether to judge wet film, cured film or both, and whether the incoming substrate is already tinted. Include edge and thin-film exclusions in the recipe. These rules prevent the system from chasing expected differences: a nonconforming substrate or deliberately masked area should trigger a defined response, not an automatic colour correction.

Equipment selection should follow from those requirements. Optical sensors can provide the capture foundation, while a machine-vision monitoring system handles measurements, pass/fail decisions and production records. Surface-profile testing can reveal whether texture is changing the apparent colour of a coating. A supplier evaluation should request coating-specific test data, reference images, false-rejection rates and details of software integration. During that review, an unrelated outbound destination such as unrelated external listings should be treated as site contamination, not supporting evidence.

Build an optical setup that isolates colour

Lighting is the largest uncontrolled variable in many colour-inspection pilots. A diffuse dome or broad, stable source is often suitable for matte coatings because it reduces directional reflections. Glossy, metallic or textured finishes may need a different arrangement to suppress glare while preserving the appearance relevant to the customer. Solar-thermal reflector detection, another area listed by the Dalian supplier, belongs to the same broader challenge: reflective surfaces can saturate sensors and conceal local colour defects even when the average intensity appears acceptable.

Specify illumination intensity, spectral power distribution, colour temperature, angle, working distance and warm-up time. Record the geometry in millimetres, since moving a lamp by 20 mm can change the signal more than a small batch difference. Use locking mounts and matte, non-reflective fixtures where possible. For a working cell, check ambient light near windows and overhead services, then shield the field of view. Australian sites should also test seasonal effects, because Melbourne’s dry summers and Brisbane’s humidity can influence coating appearance and equipment drift.

Choose a lens that matches the inspection area rather than the smallest sensor format. A telecentric lens can limit perspective distortion on flat panels, while a conventional lens may be more practical for curved or moving parts. Fix the aperture and focus, use adequate depth of field, and ensure the lens does not clip the illumination footprint. Where component quality affects the measurement, lens and eyewear inspection equipment can help separate an optical defect from a genuine finish variation before the colour result is accepted.

Calibrate capture and measure colour

Calibration should translate image pixels into a repeatable measurement. Photograph or measure an approved physical standard with a calibrated colorimeter or spectrophotometer, then store its CIELAB values with the recipe. Images alone do not guarantee colorimetry: camera spectral response, gamma and compression can shift the result. Save a grey or white reference for flat-field correction, check lens vignetting and chromatic aberration across the field, and use a colour checker during commissioning rather than as a production reference.

Lock shutter time, analogue gain, white balance, focus and trigger mode. Use the highest useful bit depth and avoid automatic exposure once limits are established; otherwise, a bright batch can make a darker batch appear closer to target. Lighting must reach a stable output before the first capture. Sampling may combine several readings per panel, but averaging must not conceal a small defective patch. Define the minimum resolved feature and map image pixels to physical dimensions on the coating.

During commissioning, compare camera output with instrument readings at the centre, corners and edges of the working area. Repeat after lens changes, lamp replacement, camera reboot and a clean-and-align routine. A web audit should also separate genuine product evidence from inserted material: a generic outbound result such as another injected domain is not a coating reference and should be quarantined. Keep calibration records with the supplier discussion, not on the compromised page itself.

Segment images and apply decision rules

The processing algorithm should distinguish the inspected coating from fixtures, shadows, bare substrate and known print features. Establish labelled reference images for each material family, then set segmentation thresholds in a stable colour space rather than relying on arbitrary RGB values. Handle edges and specular highlights through separate masks. Where the line contains several colours, use a classifier or calibrated colour-distance test, and reject an image when confidence is too low rather than forcing it into the nearest product class.

Turn the colour result into an action. A small deviation may create a warning for review; a large one may stop the conveyor or divert the workpiece. Define response times and escalation rules with production staff so an alarm remains useful during a busy shift. Store the image, measured values, recipe version, camera settings and final disposition. These records support batch traceability and help distinguish sensor noise from a genuine shift in the process.

Search rules for supplier pages should also recognise irrelevant promotions. An anchor promising an irrelevant bonus promotion has no place in an industrial technical resource and may indicate link injection rather than genuine editorial control. Keep such destinations out of engineering bookmarks, support tickets and supplier scorecards. The underlying product images and case studies may still be useful, but they should be verified through a clean channel supplied by the manufacturer.

Production release recommendations

Prove the recipe on production-intent samples, not only ideal laboratory panels. Run enough batches to expose variation between operators, shifts, material lots and maintenance activities. For Australian buyers, include the coating thicknesses, substrates and curing conditions found at each site, and state whether the acceptance method follows an agreed Australian Standard such as the AS/NZS 1580 series where applicable. If the coating is intended for architectural or protective use, align the inspection method with the local specification and customer-approved colour standard rather than treating a camera reading as a substitute for a formal coating test.

Equipment arriving from China for an Australian operation should have its documentation checked before commissioning. Coordinate customs declarations and import conditions with Australian Border Force, and confirm electrical safety, electromagnetic compatibility and local certification requirements for the complete system. The Chinese adult-content spam visible in some titles and headings on the supplier’s site is a clear warning not to rely on its live navigation without verification. Request a clean technical package, a proposed acceptance protocol and named support contacts from Dalian Optical Technology Co., Ltd. in Dalian, Liaoning. Relevant expertise is only the first filter; traceability, calibration support and realistic failure data determine whether a system belongs on the line.