How Do Vision Systems Handle Color Variations in Products?

Colour inspection is possible, but colour is more sensitive than many factories expect.

A product can look different under different lights. A camera can see colour differently from the human eye. A glossy surface can reflect surrounding light. A supplier batch can shift slightly. Packaging artwork may vary between print runs.

Computer vision can handle colour variations, but only when lighting, camera settings, calibration, tolerance rules, and product context are controlled.

Colour inspection is not just "teach the system red and green." It is a measurement problem.

Why colour varies in production

Colour can vary for many reasons:

• Raw material batch variation
• Paint, coating, or dye differences
• Printing process variation
• Curing or drying differences
• Surface texture
• Gloss or reflection
• Dust or contamination
• Ageing or fading
• Lighting changes
• Camera exposure changes

Some variation is acceptable. Some indicates a defect. The vision system must be taught the difference.

Lighting control is the first requirement

Colour inspection without controlled lighting is unreliable.

If a product is inspected under changing ambient light, the camera may report colour differences that are not real product differences. Sunlight, overhead lights, shadows, and reflections can all affect the image.

A good colour inspection setup uses stable, controlled lighting and shields the inspection area from external variation where possible.

This may include enclosed inspection stations, diffuse lighting, polarisation, fixed exposure, and regular calibration.

Camera settings must be locked

Automatic camera settings can create problems. Auto exposure, auto white balance, and auto gain may change the image between inspections. That makes colour comparison unstable.

For manufacturing inspection, settings should usually be fixed after calibration:

• Exposure
• Gain
• White balance
• Focus
• Lighting intensity
• Working distance

The goal is repeatability. The same product under the same condition should produce the same image.

Define acceptable tolerance, not perfect colour

Factories rarely need every product to match a theoretical perfect colour exactly. They need to know whether the colour is within acceptable tolerance.

Tolerance may depend on customer standard, brand requirement, functional need, or internal quality rule.

For example, a slight shade variation may be acceptable for an internal component but unacceptable for visible packaging. A colour-coded safety part may need tighter control than a hidden bracket.

The vision system should be configured around business and quality requirements, not arbitrary sensitivity.

Colour spaces and calibration matter

Vision systems may use colour spaces such as RGB, HSV, Lab, or other transformed representations depending on the application.

For serious colour work, calibration targets and reference samples may be used. The system may compare inspected parts against known acceptable samples rather than relying on raw camera values.

This helps reduce false rejections caused by camera or lighting differences.

Handling product families and variants

If a line runs multiple colours or variants, the system must know which colour is expected for the current SKU or batch.

This can be handled through:

• Operator recipe selection
• Barcode or label recognition
• ERP or MES work order integration
• Product classification
• Variant-specific tolerance bands

Connected production context matters. AICAN Optiwise can help align inspection with the active product, batch, and order so the system does not reject a correct colour because it is using the wrong recipe.

Reflection and gloss create special challenges

Glossy products can make colour inspection difficult because the camera may capture reflected light instead of true surface colour.

Solutions may include:

• Diffuse lighting
• Dome lighting
• Polarising filters
• Controlled camera angle
• Shielding from ambient light
• Multiple images under different lighting

The right approach depends on material and surface finish.

AI can help, but calibration still matters

AI-based vision can sometimes handle natural variation better than fixed thresholds, especially when colour variation appears with texture, print defects, stains, or surface irregularities.

But AI does not remove the need for stable imaging. If lighting changes randomly, the model may learn the wrong patterns or become unreliable.

For colour-sensitive inspection, good image discipline comes first.

What should be validated before deployment?

Use real production samples:

• Known acceptable colours
• Known unacceptable colours
• Borderline cases
• Different supplier batches
• Different print runs
• Parts under normal production handling
• Product variants
• Samples after cleaning or curing where relevant

Validation should test whether the system rejects only what the quality team agrees is unacceptable.

Where AICAN Optiwise fits

AICAN Optiwise helps manufacturers connect colour inspection results with production, batch, supplier, and quality context. If colour variation increases for one material lot or one line, teams can see the pattern faster.

AICAN focuses on making factory data actionable, not isolated. You can learn more at About AICAN.

Founder's Note

Colour is one of those inspection problems where humans say, "I can see it," but systems need discipline to see it the same way every time. The answer is not only better software. It is controlled lighting, agreed tolerances, and clear product context.

When those are in place, colour inspection becomes far more reliable.

FAQs

1. Can computer vision check exact colour match?

It can check colour within defined tolerance, but exact colour measurement requires controlled lighting, calibration, and agreed quality standards.

2. Why does lighting affect colour inspection so much?

Because cameras capture reflected light. If the light source changes, the captured colour can change even if the product has not changed.

3. Can vision systems handle multiple product colours?

Yes, with recipes, product identification, or integration to the active work order so the expected colour is known.

4. Are glossy products harder to inspect?

Yes. Gloss creates reflections that can confuse colour readings. Diffuse lighting, filters, and controlled angles can help.

5. Can Optiwise help with colour variation analysis?

Optiwise can help connect inspection results to batch, supplier, production, and quality data so recurring colour variation patterns are easier to investigate.