SWIR Camera Lens for Contamination Detection: How to Find Foreign Material That Matches the Product in Visible Light

One of the hardest problems in automated quality inspection occurs when contamination does not look like contamination. A foreign particle may have almost the same colour as the product, a fragment of unwanted polymer may visually resemble the correct material, an organic contaminant may disappear against a similarly coloured food surface, or a foreign fiber may be impossible to distinguish from the surrounding product using an ordinary grayscale or RGB camera. In these cases, increasing visible-light resolution does not necessarily solve the problem because the fundamental limitation is not the number of pixels. The limitation is lack of material contrast.

A SWIR camera lens for contamination detection allows an industrial imaging system to work within approximately 900–1700 nm, where materials can exhibit wavelength-dependent differences in absorption and reflectance that are unrelated to visible colour. This creates an important opportunity: two objects that produce nearly identical visible images can sometimes produce substantially different SWIR responses because their material composition is different. Scientific research comparing visible and SWIR imaging has demonstrated significantly stronger discrimination between visually similar materials when short-wave infrared information is used.

For manufacturers, the practical question is therefore not simply whether foreign material is visible. It is whether the contaminant produces enough SWIR contrast against the host product to be detected reliably at the required particle size and production speed. The dedicated Kyptec Automation® SWIR Camera Lens collection currently includes five focal lengths—8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm—giving machine builders several ways to balance inspection width, working distance and pixels available on the contaminant. The current range is listed around 900–1700 nm operation, 2 MP resolution, 2/3-inch sensor format, F1.4 aperture and C-Mount.

Why Foreign Material Can Disappear in Visible-Light Inspection

Visible inspection primarily responds to colour, brightness, surface texture and geometry. If the contaminant and acceptable product have similar values for these properties, segmentation becomes difficult.

Imagine a light-coloured foreign polymer fragment lying on a light-coloured food product. If both reflect similar amounts of visible light, their pixel intensities may overlap. Increasing image resolution can define their edges more sharply, but it does not necessarily create a stronger difference between the two materials.

SWIR imaging introduces another dimension: material-dependent optical behaviour. Two visually similar objects can absorb or reflect short-wave infrared wavelengths differently because their molecular structures, moisture content or chemical composition differ. Research on SWIR material classification has shown that this spectral information can substantially improve separation of objects that are ambiguous in the visible spectrum.

For contamination inspection, this means the machine can potentially distinguish “correct material” from “wrong material” rather than merely “light object” from “dark object.”

Contamination Detection Begins With Host-to-Contaminant Contrast

The most important feasibility test is not whether the contaminant produces an interesting SWIR spectrum in isolation. It is whether its response is sufficiently different from the host product under the final inspection geometry.

Let the measured response of the acceptable product be (P(λ)) and the contaminant response be (C(λ)). A simple normalized difference at a selected wavelength can be represented as:

D(λ) = |C(λ) − P(λ)| / [C(λ) + P(λ)]

A larger stable value of (D) generally gives the classification system more separation to work with.

However, the strongest wavelength on one perfect laboratory sample should not automatically become the production wavelength. The comparison should include normal variation in the host material and contaminant population.

The real goal is:

minimum contaminant-to-product separation under maximum normal production variation.

That distinction is central to a reliable foreign-material detection system.

The Best SWIR Contaminant Is Not Necessarily the Most Visually Different One

A foreign object that looks dramatically different in visible light may not need SWIR inspection at all. Conventional machine vision could already detect it effectively.

SWIR becomes particularly valuable when the contaminant is spectrally different but visually similar.

Examples may include selected polymer fragments, organic foreign material, fibers, ingredients with incorrect composition, residues or material substitutions where colour inspection alone provides poor confidence.

This is an important purchasing consideration. SWIR should be selected because the required defect has a material-information problem, not simply because infrared imaging appears more advanced.

Contamination Detection Is Different From General Material Identification

Material identification asks, “What material is this object?”

Contamination detection asks a more operational production question:

“Is there any unacceptable material inside this field that should not be present?”

That distinction changes system design.

A material-classification station may classify a complete object from its average response. A contamination detector must often find a very small region inside a much larger acceptable product.

The contaminant can therefore occupy only a tiny percentage of the image.

This makes spatial sampling far more important.

A system capable of identifying a 30 mm polymer piece does not automatically have enough spatial resolution to detect a 1 mm fragment of the same polymer.

Contaminant-to-Host Area Ratio Determines Signal Purity

Suppose a region of interest contains 100 pixels.

If the contaminant occupies 80 pixels, its spectral response dominates.

If it occupies 5 pixels, the measured region is overwhelmingly influenced by the host material.

If it occupies less than one effective resolved pixel, its optical signal may be mixed almost completely with the surrounding product.

This creates one of the most important rules in SWIR contamination inspection:

The contaminant must occupy enough spatially independent pixels to preserve its material identity.

The exact number cannot be universal because optical blur, noise and classifier design vary, but relying on one barely resolved pixel is rarely a strong industrial strategy.

Mixed Pixels Can Make Foreign Material Look Like Acceptable Product

At a contaminant boundary, one pixel can receive radiation from both the contaminant and the host material.

A simplified mixed response can be described as:

M(λ) = αC(λ) + (1 − α)P(λ)

where (α) represents the fraction of the pixel occupied by contamination.

If (α) is small, the measured spectrum moves toward the host material.

This explains why tiny particles can become difficult even when pure samples show excellent SWIR separation.

The engineering solution is not simply “use a more sensitive threshold.” The stronger approach is to design field of view and working distance so the smallest required contaminant occupies enough pixels to create a stable material response.

Calculate Pixels per Contaminant Before Selecting the Lens

Assume the horizontal sensor contains 1600 pixels and the required inspection field is 400 mm wide.

Object-side sampling is approximately:

400 ÷ 1600 = 0.25 mm/pixel

A 2 mm contaminant spans approximately:

2 ÷ 0.25 = 8 pixels

before considering blur and boundary mixing.

If the same sensor covers an 800 mm field:

800 ÷ 1600 = 0.50 mm/pixel

The same 2 mm contaminant spans only about four pixels.

This is why contamination detection often benefits from a tighter FOV than broad material classification. The machine must cover the production width while preserving enough pixels on the smallest foreign material that must trigger rejection.

Contaminant Size Should Be Defined From the Reject Specification

Statements such as “detect small contamination” are not sufficient for lens selection.

The buyer should specify something measurable:

“Detect foreign material ≥2 mm.”

“Reject contamination area ≥4 mm².”

“Detect elongated foreign fibers ≥5 mm long and ≥0.5 mm wide.”

“Detect foreign fragments occupying at least X pixels under the validated setup.”

Only after the rejectable defect is defined can the camera field, focal length and working distance be engineered properly.

This specification also prevents the project from being judged against particles substantially smaller than the optical design was intended to resolve.

SWIR Contrast Must Be Validated Across Real Contamination Types

A production line rarely encounters one perfect foreign-material sample.

The contamination population may contain different:

sizes;

thicknesses;

orientations;

suppliers;

colours;

surface textures;

moisture conditions;

or chemical compositions.

A robust feasibility study therefore requires multiple samples from every relevant contamination type.

If the system is intended to detect “foreign polymer,” the testing should not rely on one clean polymer chip. It should include the real polymer variations likely to reach production.

The objective is to establish whether all commercially important contaminant populations remain sufficiently separate from acceptable product.

Same Colour Does Not Mean Same SWIR Signature

This is one of the most valuable properties of SWIR contamination inspection.

A dark foreign object and a dark product can look nearly identical in visible light, yet their SWIR reflectance curves may differ.

Conversely, two objects with very different visible colours may show similar SWIR responses in part of the spectrum.

Therefore, visible appearance should not be used to predict spectral separability.

The application should be tested directly across the usable 900–1700 nm region.

This is especially important for manufacturers trying to detect foreign material that repeatedly escapes conventional colour cameras.

Moisture Can Either Improve or Complicate Contamination Contrast

Water-related absorption within the SWIR range can create strong differences between materials with different moisture contents.

This can be useful when a contaminant naturally contains substantially more or less moisture than the host product.

However, host-product moisture itself may vary during production.

If an acceptable product becomes wetter, its SWIR response can move toward the contaminant distribution and create false positives.

The correct validation set should therefore include the minimum and maximum acceptable host moisture level.

A contamination system should distinguish foreign material from normal product variation, not merely from one nominal reference sample.

Product Temperature, Surface Texture and Orientation Can Create Nuisance Variation

Industrial samples are not presented as perfectly identical laboratory targets.

A rough surface scatters differently from a smooth one. A curved object changes illumination angle. A rotating product can expose different facets. A hot or cold process may change the measurement environment.

These variations can broaden the normal-product distribution.

A strong SWIR inspection system should therefore be designed around the worst realistic host variation.

The useful performance metric is not how strongly a contaminant separates from one perfect product image. It is how strongly it separates from the complete population of acceptable products.

Small Contaminants Need More Than Spectral Contrast

SWIR cannot replace spatial resolution.

A contaminant may have an extremely distinct spectrum, but if it occupies a fraction of a pixel its signal can be diluted by surrounding material.

Conversely, a large contaminant with only moderate spectral contrast may be easy to classify because it occupies many pure pixels.

This means contamination detectability depends on both dimensions:

spectral separability × spatial separability.

A strong lens choice preserves both.

Kyptec Automation® KL-1408 for Broad Foreign-Material Screening

For applications where a wide conveyor or broad product stream must be inspected and the minimum foreign material is relatively large, the Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens can be evaluated as the widest focal-length option in the current portfolio. The live model is specified for 8.5 mm focal length, 900–1700 nm, 2 MP, F1.4, 2/3-inch format and C-Mount.

Its broad field can be useful for first-stage screening where throughput and production width are major requirements. The trade-off is lower pixels per millimetre compared with a tighter field, so the smallest contamination target should be validated at the actual working distance before choosing a wide configuration.

Kyptec Automation® KL-1410 for Balancing Coverage and Contaminant Sampling

When the field does not need to be extremely wide, the Kyptec Automation® KL-1410 12.5 MM SWIR Camera Lens provides an intermediate option.

Its current specifications include 12.5 mm focal length, 900–1700 nm operation, 2 MP resolution, F1.4 aperture, 2/3-inch format and C-Mount.

For a contamination machine, reducing unused background and concentrating the sensor more effectively on the product can provide more pixels on each foreign particle without moving immediately to a very narrow inspection field.

This makes intermediate focal lengths useful when both conveyor coverage and small-contaminant detection matter.

A Contamination Mask Should Separate Product From Background First

Foreign-material classification becomes more stable when the software first determines where valid product exists.

If background pixels, conveyor pixels and empty space are included in the contamination classifier, they add unnecessary variability.

A useful pipeline is conceptually:

identify product region → exclude background → evaluate material consistency → locate abnormal material region → apply size and confidence rules.

This is especially important for loose products or irregular shapes because product edges can otherwise create false contamination signals.

The background material itself should also be selected so its SWIR response is stable and clearly different from the product where possible.

Edge Pixels Should Often Be Treated Differently

Product boundaries naturally contain mixed pixels.

A pixel can contain partly product and partly conveyor, producing a spectral response that matches neither.

If the system interprets every unusual edge response as foreign material, false rejection can increase substantially.

One practical strategy is to use an edge exclusion or special boundary model while retaining full contamination sensitivity in the interior.

The required exclusion width should remain smaller than the smallest legitimate contamination zone that must be detected near an edge.

This is a more robust strategy than simply lowering global sensitivity to suppress edge false positives.

Foreign Material Near an Existing Product Boundary Is Harder

A contaminant lying completely inside a uniform product region provides many opportunities for clean host-versus-contaminant comparison.

A contaminant touching the product boundary is harder because its pixels can be mixed with background as well as product.

The qualification set should therefore include contaminants:

in the center;

near edges;

near corners;

partly overlapping a boundary;

and in geometrically difficult locations.

A system demonstrated only with centrally positioned defects may substantially overstate real production capability.

Contaminant Thickness Can Change Its Visibility

A thick contaminant generally contributes more of its own optical response than an extremely thin film.

Thin residue can combine spectrally with the surface underneath it, producing a mixed signature rather than a pure contaminant signature.

This is why foreign-particle detection and residue-film detection should not automatically share the same acceptance threshold.

The buyer should define whether the target is:

a discrete particle;

a fiber;

a thin film;

a smear;

a layer;

or a material inclusion.

Each presents a different optical problem.

Kyptec Automation® KL-1412 for Smaller Foreign Material in Controlled Inspection Zones

Where the machine can use a narrower field, the Kyptec Automation® KL-1412 25 MM SWIR Camera Lens can allocate more of the available sensor area to the product. The live product page specifies 25 mm focal length, 900–1700 nm operation, 2 MP resolution, F1.4 aperture, 2/3-inch format and C-Mount.

This can be valuable for individual-product inspection, targeted contamination analysis and controlled conveyor lanes where small foreign regions must occupy more pixels.

A narrower FOV does not create better spectral contrast by itself. Its advantage is preserving more spatial information from the contaminant once useful SWIR material contrast has already been established.

F1.4 Can Help When Short Exposure Is Required

Fast production lines reduce the exposure time available before motion blur becomes significant.

For example, if a conveyor moves at 2 m/s and exposure is 250 µs:

2000 mm/s × 0.00025 s = 0.5 mm motion during exposure.

For a 2 mm contaminant, 0.5 mm of motion is significant.

Shorter exposure reduces blur but also reduces collected SWIR energy.

The F1.4 maximum aperture used across the Kyptec Automation® SWIR lens portfolio can therefore provide useful light-gathering capability in high-throughput contamination inspection.

However, aperture should still be balanced against required depth of field when product height varies.

Motion Blur Can Dilute the Contaminant Signature

Motion blur has a particularly damaging effect on material-sensitive defect inspection.

A small foreign particle that moves across several pixels during exposure spreads its signal into surrounding acceptable-product pixels.

Instead of obtaining several strong contaminant pixels, the system obtains a larger group of weak mixed pixels.

This can push the contaminant toward the accepted-product distribution.

The correct exposure limit should therefore be established from classification reliability on the minimum contamination size, not merely from whether the moving image looks subjectively sharp.

Detection Thresholds Should Include Both Material Confidence and Defect Area

Rejecting every unusual pixel can make the system unstable.

A stronger contamination decision can combine:

material abnormality confidence;

contiguous abnormal area;

minimum number of pixels;

shape or persistence where appropriate;

and location within the valid product region.

For example, one isolated low-confidence abnormal pixel may be ignored, while ten adjacent pixels with strong foreign-material probability trigger rejection.

This approach uses both spectral and spatial evidence.

The exact thresholds must be determined from real production validation.

False Positives Must Be Built Into the Development Dataset

Foreign-material detection systems often fail commercially because the development set contains many contamination examples but too few difficult good products.

Examples of false-positive challenges may include:

dark edges;

surface folds;

natural composition variation;

wet spots;

texture changes;

specular regions;

small holes;

acceptable inclusions;

or production marks.

Every recurring good-product condition that initially triggers the system should become part of the validation dataset.

Over time, the most useful dataset is not simply “good versus contaminant.” It becomes a detailed library of real product conditions that must not be confused with contamination.

Reject Thresholds Should Reflect the Cost of Escape and Waste

Extremely aggressive thresholds can reduce contaminant escape while increasing rejection of acceptable product.

Very relaxed thresholds reduce false rejects but increase contamination risk.

The correct operating point depends on the commercial consequences.

A food-quality application, high-purity material stream and non-critical cosmetic production line may require very different risk balances.

The buyer should therefore specify target:

contaminant detection probability;

false-accept rate;

false-reject rate;

minimum contaminant size;

and acceptable uncertain-product rate.

A generic statement such as “high accuracy” is insufficient for production acceptance.

Validate the Hardest Contaminant, Not the Easiest Demonstration Sample

The contaminant that looks most dramatic on a SWIR monitor should not determine system qualification.

The hardest relevant contaminant is more important.

If foreign material A separates strongly but foreign material B produces a response close to acceptable product, B determines the difficult decision boundary.

A serious qualification process should rank contaminant classes by minimum spectral separation from the host.

The production system should then be optimized around the most important difficult pair rather than the easiest visible example.

The Kyptec Automation® KL-1414 Can Support Tightly Controlled Contamination Inspection

For a smaller product region or controlled inspection cell, the Kyptec Automation® KL-1414 35 MM SWIR Camera Lens provides a tighter field. Its current specifications include 35 mm focal length, 900–1700 nm, 2 MP, F1.4, 2/3-inch format and C-Mount.

This geometry can be useful when foreign material is small and the product area itself does not require a wide scene. More sensor pixels can then be devoted to the contamination-sensitive region.

Such a design is often more useful than capturing large amounts of irrelevant background simply to maximize FOV.

Unknown Foreign Material Needs a Safe Decision Policy

A production machine may encounter contamination that was never included during development.

If the classifier only knows “acceptable product” plus a few named contaminant classes, an unseen material can be incorrectly assigned to the closest known category.

For many contamination applications, the better philosophy is not necessarily to identify every foreign substance precisely.

It is to determine whether a pixel or region remains consistent with acceptable product.

Anything sufficiently outside the validated acceptable population can then be rejected or classified as unknown contamination.

This open-set approach can be valuable where unpredictable foreign material is possible.

Contamination Detection Should Often Be Designed Around “Product Purity”

Instead of teaching the machine every possible unwanted material, certain applications are better framed as purity inspection.

The system establishes the normal SWIR distribution of acceptable product across approved lots, suppliers and process conditions.

Regions that fall outside that validated distribution are treated as suspicious.

This approach is particularly useful when the number of possible contaminant types is very large.

It also changes the procurement question from:

“Can the system recognize twenty contaminants?”

to:

“Can the system reliably distinguish acceptable product from any materially abnormal region large enough to matter?”

That is often a stronger production objective.

Why Kyptec Automation® Is a Strong Platform for SWIR Contamination Detection

The Kyptec Automation® SWIR Camera Lens collection currently provides five focal lengths—8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm—giving OEMs flexibility to design contamination inspection according to actual conveyor width, working distance and minimum defect size. The collection currently lists five products, while individual product pages specify the shared 900–1700 nm wavelength range, 2 MP resolution class, F1.4 aperture, 2/3-inch sensor format and C-Mount.

This range is useful because contamination applications can vary considerably. Broad screening can require wide coverage, while small-particle detection may require much tighter framing. Kyptec Automation® provides a focused SWIR optical family that lets machine builders make that geometry decision without moving away from the same specialized short-wave infrared lens category.

Frequently Asked Questions About SWIR Camera Lenses for Contamination Detection

1. Why does foreign material sometimes disappear completely in a normal camera image?

A visible camera separates objects mainly through visible colour, brightness, texture and shape. When the contaminant and product have similar visible reflectance, there may be very little image contrast even if their compositions are completely different. SWIR can potentially separate them because materials that appear similar in visible light can exhibit different short-wave infrared reflectance characteristics.

2. Does a foreign material need to be a different colour for SWIR detection to work?

No. One of the main reasons to evaluate SWIR is precisely when visible colour does not provide sufficient separation. A contaminant can closely match the host product in colour while having a different SWIR response because of differences in material composition. Feasibility should be based on measured SWIR contrast rather than visual appearance.

3. What is more important for contamination detection: spectral difference or particle size?

Both are essential. Strong spectral separation is of limited value when the particle is too small to occupy enough resolved pixels, while a large contaminant can remain difficult if its SWIR response closely matches the product. Production detection capability is therefore determined by the combination of material contrast and spatial sampling.

4. Can SWIR detect contamination that occupies only one pixel?

A one-pixel event is generally a weak basis for robust industrial classification because blur, noise and mixing with surrounding product can dominate the measurement. It is preferable for the minimum rejectable contaminant to occupy several meaningful pixels so both material response and spatial continuity can support the decision.

5. Why does a tiny contaminant sometimes have a weaker SWIR signature than the same material in a large sample?

Small particles generate mixed pixels. The recorded signal contains contributions from both contaminant and surrounding product, causing the measured spectrum to move toward the host-material response. A larger sample produces more pure contaminant pixels and therefore a stronger material signature.

6. How should I specify minimum foreign-material size when requesting a SWIR inspection system?

Use measurable dimensions rather than terms such as “small particle.” Specify minimum width, length, diameter or area, and state whether fibers, films or irregular fragments are included. Also provide required inspection width and working distance because these values determine pixels per millimetre and therefore whether the minimum contaminant can be spatially resolved.

7. Can a thin contamination film be detected as easily as a solid foreign particle?

Usually not. A thin film can allow the substrate response to remain dominant, creating a mixed spectral signature. A thicker particle contributes more strongly to the recorded signal. Thin residue should therefore be validated separately from discrete foreign objects, even when both consist of the same material.

8. Can acceptable product variation cause false contamination alarms?

Yes. Moisture, composition, texture, thickness and surface geometry can change the normal SWIR response. If development uses only one ideal good sample, legitimate production variation can later appear abnormal. The acceptable-product library should therefore contain representative manufacturing extremes.

9. How can I reduce false contamination detection around product edges?

Product boundaries naturally contain mixed pixels containing product and background. A dedicated edge model, boundary exclusion zone or segmentation strategy can reduce these false positives without weakening sensitivity across the product interior. Edge-located contamination must still be included during qualification so the exclusion strategy does not create an inspection blind zone.

10. Should a SWIR contamination system identify the exact foreign material or simply reject it?

That depends on the production objective. Many applications do not need exact contaminant identity; they only need to determine that the material is inconsistent with acceptable product. A product-purity or anomaly-detection strategy can be more practical when many unknown foreign materials are possible.

11. How should conveyor background be selected for SWIR contamination inspection?

The background should ideally have a stable SWIR response that separates clearly from acceptable product and likely contaminants. It should also resist changes caused by wear, dirt and moisture. Background samples should be included in validation because mixed pixels near product edges can influence the contamination classifier.

12. Which Kyptec Automation® SWIR lens should be evaluated for a wide contamination-inspection conveyor?

The Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens provides the widest focal-length geometry in the current family and can be considered when broad coverage is the priority. The minimum contaminant should still be checked in pixels at the actual field of view because excessively wide coverage can reduce small-particle detectability.

13. When does the Kyptec Automation® KL-1412 25 MM SWIR Camera Lens become useful for contamination inspection?

The Kyptec Automation® KL-1412 25 MM SWIR Camera Lens is useful to evaluate when the field can be narrower and smaller contamination regions need stronger pixel representation. It can suit controlled individual-product or narrower-lane inspection where maximum conveyor width is not the primary requirement.

14. Can increasing exposure make small contamination easier to detect?

Longer exposure can improve signal when illumination is limited, but it can also increase motion blur on a moving production line. For small foreign material, blur can dilute the contaminant signal into surrounding product pixels. Exposure therefore has to be optimized from signal-to-noise ratio and contaminant motion together.

15. How should I test a contamination detector before approving it for production?

Use independently verified good products plus each relevant contaminant at multiple sizes, locations and orientations. Include contaminants at the minimum rejectable size, difficult product edges, normal host variability and high production speed. Acceptance should be based on measured false accepts and false rejects rather than a small number of demonstration images.

16. What should happen when a foreign region has low classification confidence?

A well-designed system can use an uncertain or review category instead of forcing every region into acceptable or contaminant classes. Products containing sufficiently suspicious low-confidence regions can be diverted for secondary inspection. This is particularly useful when unknown foreign materials may occur.

17. Is contamination detection better if I train on as many foreign materials as possible?

Not necessarily. Dataset diversity is useful, but adding many poorly represented contaminant classes can make the decision boundary complicated. In some applications, building a strong definition of normal product and detecting deviations from that population is more robust than attempting to name every possible contaminant.

18. What information should I provide when buying a SWIR camera lens for contamination detection?

Provide product dimensions, inspection width, camera sensor size, working distance, production speed, minimum contaminant dimensions, contaminant materials, acceptable product variations and whether the objective is exact contaminant classification or general foreign-material rejection. These values allow focal length and FOV to be selected around the actual detection requirement instead of choosing optics from focal length alone.

19. How can I determine whether my contamination problem is suitable for 900–1700 nm SWIR inspection?

Test representative host material and all critical contaminants across the usable SWIR range under controlled illumination. Compare not just perfect samples but normal production variation. If the hardest contaminant remains sufficiently separated from the complete acceptable-product population and can be spatially resolved at the required FOV, the application has a much stronger basis for industrial automation.

20. Why is Kyptec Automation® a strong choice for SWIR foreign-material detection?

Kyptec Automation® offers a dedicated SWIR Camera Lens collection covering five focal lengths from 8.5 mm through 50 mm. The current product family is designed around 900–1700 nm imaging with 2 MP resolution class, 2/3-inch sensor format, F1.4 aperture and C-Mount. This gives OEMs useful optical flexibility to move from broad contamination screening to tightly framed foreign-material inspection while staying within one specialized Kyptec Automation® SWIR Camera Lens platform.

Conclusion

SWIR contamination detection is most valuable when foreign material is difficult to distinguish by appearance but different in composition. A contaminant does not have to be brighter, darker or differently coloured than the product. It only needs to produce a sufficiently different and repeatable SWIR response within the wavelengths available to the imaging system. Research has demonstrated that short-wave infrared material information can substantially improve differentiation between visually similar objects where visible-spectrum information is ambiguous.

The engineering challenge is that spectral difference alone is not enough. A foreign particle must occupy enough pixels to preserve its material signature, and its contrast must remain distinct from the complete variation of acceptable production. Small contaminants create mixed pixels, thin films combine with the substrate response, motion blur can dilute foreign-material information, surface variation can create false positives, and excessively wide fields can reduce a highly distinctive contaminant to only a few unreliable pixels. This is why the strongest inspection systems connect material contrast, minimum contamination size, FOV, working distance, exposure, production speed and rejection thresholds before the final lens is selected.

For an OEM or industrial buyer, the correct development sequence is to begin with the hardest contamination pair. Identify the foreign material whose SWIR response is closest to acceptable product. Measure that separation across normal product moisture, colour, texture, thickness and supplier variation. Then define the smallest contamination size that must trigger rejection. From that requirement, calculate the maximum permissible object-side pixel size and therefore the field of view that the optical system can support. Only after these limits are known should focal length be selected.

The Kyptec Automation® SWIR Camera Lens collection provides a strong foundation for this process because its five focal lengths allow inspection geometry to be adapted without moving away from the dedicated SWIR optical family. A broad production stream can begin with the wide coverage available from the Kyptec Automation® KL-1408; intermediate requirements can use a tighter field to improve contaminant sampling; and controlled inspection regions can move toward the Kyptec Automation® KL-1412 or Kyptec Automation® KL-1414 when more pixels are required across the foreign material. The common 900–1700 nm orientation, F1.4 aperture, 2 MP resolution class, 2/3-inch format and C-Mount provide a practical optical basis for OEMs developing automated foreign-material inspection systems.

The strongest design principle is therefore straightforward: do not ask whether the contaminant can be seen—ask whether it can be separated materially, spatially and statistically from everything that is allowed to pass. When the contaminant produces useful SWIR contrast, occupies enough pixels, remains distinguishable from acceptable variation and is validated under full production conditions, a Kyptec Automation® SWIR Camera Lens can provide a technically strong optical platform for finding foreign material that ordinary visible inspection may fail to distinguish.