SWIR Camera Lens for Textile and Fiber Inspection: Material Identification, Moisture Variation, Contamination and Fabric Sorting

Textile inspection becomes significantly more difficult when quality depends on what the fabric is made from rather than how it looks. Cotton and polyester fabrics can be dyed to nearly identical colours, recycled garments may contain unknown fiber blends, residual moisture can vary across a fabric web without producing an obvious visible mark, and oils, finishes or foreign fibers can create material differences that conventional RGB inspection does not consistently distinguish. A SWIR camera lens for textile and fiber inspection helps move the inspection process from appearance-based judgment toward material-sensitive imaging by allowing a compatible SWIR camera to record wavelength-dependent absorption and reflectance information across approximately 900–1700 nm.

This spectral region is highly relevant to textile identification because natural, regenerated and synthetic fibers contain different molecular structures. Cellulosic fibers, polyester-based materials, polyamides, protein fibers and elastomer-containing blends do not interact identically with near-infrared and short-wave infrared radiation. Published research has demonstrated that NIR/SWIR imaging can differentiate important textile classes, including cotton and polyester, estimate polyester concentration in blended fabrics and support automated sorting of textile waste. Research using approximately 1300–1700 nm has shown clear differentiation between cotton and polyester, while broader work using 900–1700 nm has demonstrated real-time classification of natural, synthetic and blended textiles.

For an OEM, however, useful textile classification requires more than selecting a camera capable of seeing infrared radiation. The SWIR camera lens must transmit the wavelengths carrying the fiber information, cover the required fabric width, maintain sufficient spatial sampling for contamination or blend boundaries, and deliver enough optical signal at production speed. The Kyptec Automation® SWIR Camera Lens collection provides 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm focal-length options within a common 900–1700 nm, 2 MP, 2/3-inch, F1.4 and C-Mount architecture. Textile is also explicitly identified among the published application areas of the current range.

Why Fiber Identification Is Different From Ordinary Fabric Inspection

Visible machine vision is extremely effective for holes, tears, colour variation, weaving defects, stains and obvious surface contamination. Those are primarily spatial or visible-reflectance problems. Fiber identification asks a fundamentally different question: is this cotton, polyester, nylon-like material, regenerated cellulose, wool, a blend, or an unknown textile composition?

That decision requires information related to material chemistry.

Natural cellulose-based fibers contain O–H and C–O related structures, while many synthetic polymers contain different C–H and other molecular combinations. These chemical groups produce characteristic wavelength-dependent absorption behaviour in NIR/SWIR imaging. Research on textile identification describes the approximately 780–1700 nm region as useful for differentiating fibers because these molecular groups influence infrared reflectance differently.

The camera therefore does not classify a fabric because cotton “looks darker” than polyester universally. It classifies the shape of the spectral response across selected wavelengths.

That distinction is crucial because absolute brightness changes with illumination, fabric orientation and surface texture, while a properly developed spectral signature can remain much more closely related to fiber composition.

Cotton and Polyester Separation Is a High-Value SWIR Textile Application

Cotton and polyester are particularly important because they represent major natural and synthetic textile streams and are frequently combined in blended fabrics.

For textile recycling, incorrect separation can reduce downstream material value because a supposedly cellulose-rich stream containing substantial polyester behaves differently during recycling from a relatively pure cotton stream. Conversely, a polyester recycling process benefits from knowing how much non-polyester fiber is entering the accepted material.

Research using hyperspectral NIR imaging has demonstrated both classification of cotton versus polyester and estimation of polyester percentage in blended textiles, including visualization of spatial composition variation across fabric samples.

For an automated sorter, this creates two distinct inspection levels.

The simpler requirement is classification: cotton-rich or polyester-rich?

The more difficult requirement is quantification: approximately what proportion of polyester is present?

The second problem requires significantly stronger calibration and should not be promised merely because the first classification works.

Fabric Blends Are More Difficult Than Pure Fibers

A pure cotton textile and a pure polyester textile may produce well-separated spectral signatures. A 50/50 blend naturally falls between them, and low-percentage components can become increasingly difficult to detect.

This is especially important for recycled garments, where fabrics may contain small quantities of elastane or other fibers that materially affect recyclability even though they represent only a minor fraction of total composition. Research on textile sorting has reported that low-percentage blend components and very thin fabrics can be more difficult to classify than dominant fiber types.

An OEM should therefore define the minimum blend percentage that must influence the sorting decision.

A machine designed only to separate pure cotton from pure polyester has a substantially easier specification than one required to distinguish 95/5, 80/20 and 60/40 blends reliably.

That required composition resolution should be established before the classifier and optical station are frozen.

Visible Colour Should Not Be Allowed to Become the Fiber Classifier

A recycling system can accidentally learn that blue garments are polyester and white garments are cotton if the training library is poorly designed.

That produces impressive laboratory accuracy and poor real-world generalization.

The fiber dataset should therefore contain multiple colours of the same material, including visually similar colours across different materials. SWIR-based textile classification is valuable precisely because material chemistry can provide information beyond visible dye colour. Research into automated textile sorting has intentionally compared differently coloured samples of the same base material to evaluate whether fiber classification can remain independent of visible appearance.

The correct production question is not whether the algorithm recognizes a garment colour. It is whether it still identifies the fiber when the dye changes.

Dye Can Still Influence the SWIR Measurement

It would be incorrect to assume that dyes have no infrared effect whatsoever.

Different dyes, finishing chemicals and pigment concentrations can modify the measured spectrum. The strength of that interference depends on wavelength and textile chemistry.

This means the training database should contain the expected colour and finishing diversity of the real production stream.

However, one useful advantage of NIR/SWIR textile inspection is that certain wavelength regions can be considerably less sensitive to visible colour than ordinary RGB inspection. Research on water-spot detection in dyed fabrics found that common textile dyes had limited influence over the approximately 1265–1626 nm region used in that study, while moisture produced a measurable change and approximately 1459 nm proved especially useful for water-spot visualization.

That illustrates why wavelength selection should be based on the desired material property and nuisance variables together.

SWIR Can Map Moisture Variation Across an Entire Textile Web

Moisture monitoring is particularly important after washing, dyeing, impregnation, coating or drying.

A single-point moisture sensor measures one location. Imaging can reveal whether drying is uniform across the width of a moving fabric.

This spatial information is highly valuable because the average moisture level can appear acceptable even while one side of the web remains wetter than the other.

NIR hyperspectral research on finished textile webs has demonstrated quantitative residual-moisture monitoring and spatial visualization of drying non-uniformity, with reported prediction errors around 0.5 wt% in controlled studies. The technique has also been demonstrated on black technical textiles, showing that useful moisture information can be recovered even where visible appearance offers little guidance.

For an OEM, the useful machine output may therefore be more than “wet/dry.” It can be a cross-web moisture map used to identify dryer imbalance, coating variation or process instability.

Moisture Inspection Should Be Separated From Fiber Classification

Water strongly influences SWIR response, which means moisture can become either useful information or an unwanted variable.

If the machine is intentionally measuring residual moisture, this sensitivity is advantageous.

If the machine is attempting to classify textile composition, uncontrolled moisture can alter the fiber spectrum and potentially reduce classification accuracy.

A textile recycling sorter may therefore need training samples covering the expected humidity and moisture range, while a textile finishing line may deliberately exploit the water-sensitive response.

The two applications can use the same general SWIR optical platform, but their calibration objectives are different.

This is an important distinction when specifying the machine.

Contamination Can Be Detected When Its Material Response Differs From the Textile

Textile contamination may include oil, processing chemicals, foreign fibers, polymer fragments, labels, backing material or other unwanted matter.

SWIR can help when the contaminant produces a sufficiently different spectral signature from the base fabric. A visually transparent or similarly coloured contamination region can sometimes become much more distinguishable in infrared because its molecular composition differs.

The OEM should build a contamination library rather than treating contamination as one universal class.

Oil and a foreign polyester fiber on cotton do not produce the same spectral response. Neither does a coating residue or adhesive patch.

Each production-relevant contaminant should therefore be measured separately, including the smallest amount that must be rejected.

Textile Recycling Requires Classification at the Garment Level and the Local Material Level

A whole garment may contain several materials.

The main body can be cotton, the stitching polyester, the waistband elastane-rich, and labels or decorative sections made from additional polymers.

This creates an important decision for automated textile recycling: is the machine trying to determine the dominant composition of the whole item, or must it identify different material regions inside the same garment?

Those are different image-analysis problems.

Dominant-material sorting can average spectral information over relatively large areas.

Local blend or trim detection requires finer spatial resolution so that separate material regions remain distinguishable.

The lens field of view must therefore be selected from the material region that actually influences the sorting decision.

Textile FOV Should Be Designed From Web Width and Minimum Material Region

Suppose a SWIR camera provides 1600 horizontal pixels and the textile inspection width is 800 mm.

Nominal object-side sampling is:

800 mm ÷ 1600 pixels = 0.50 mm/pixel.

A 10 mm contamination patch spans approximately 20 pixels before lens blur and textile motion are considered.

If the same sensor is required to cover 1600 mm, sampling becomes 1 mm/pixel and the same contamination region occupies only around ten pixels.

For broad material classification this may remain sufficient. For small foreign fibers or localized finish defects it may not.

This is why a textile-machine OEM should specify both maximum fabric width and minimum meaningful defect or material patch size.

Kyptec Automation® KL-1408 for Wide Textile and Fabric Coverage

For broad textile webs or sorting conveyors where large physical coverage must be achieved from limited stand-off, the Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens provides the shortest focal length in the current range.

Its position within the dedicated 900–1700 nm Kyptec Automation® portfolio makes it useful to evaluate where a relatively wide fabric or garment-sorting area must fit inside the available 2/3-inch image format.

The important trade-off is object sampling. A wide-angle configuration should be used because the process requires the coverage, not simply because viewing more fabric appears advantageous.

If the smallest contaminant or blend region becomes too small in the image, the OEM may need a narrower FOV or multiple imaging stations.

Kyptec Automation® KL-1410 for Wide Coverage With Stronger Material Sampling

The Kyptec Automation® KL-1410 12.5 MM SWIR Camera Lens provides a useful alternative where the widest 8.5 mm geometry captures more fabric surroundings than necessary. Its current specifications include 12.5 mm focal length, 900–1700 nm operation, F1.4, 2 MP, 2/3-inch format and C-Mount.

This can be a particularly useful geometry for automated garment-sorting conveyors, narrower textile webs and material-identification stations where broad coverage remains important but individual textile regions need stronger pixel representation.

For OEM buyers, this illustrates why focal length should be selected after the usable sorting width is known rather than simply requesting a “wide-angle SWIR lens.”

Fabric Texture Can Alter Reflectance Without Changing Fiber Chemistry

Woven, knitted, brushed, nonwoven and pile textiles do not reflect SWIR radiation identically even when they contain the same fiber.

Surface roughness changes scattering. Yarn orientation changes local geometry. Folds and wrinkles create shadows.

A fiber classifier must therefore be trained across realistic textile structures rather than only flat laboratory swatches.

This is particularly important in textile recycling, where garments arrive creased and irregular.

If the same polyester fabric changes class when folded, the model has learned geometry rather than material chemistry.

Illumination and normalization should therefore be engineered to reduce orientation dependence before more algorithmic complexity is introduced.

Thin Fabrics Can Contain Background Signal

A thin or loosely woven textile may transmit enough SWIR radiation for the conveyor beneath it to influence the recorded spectrum.

This creates a mixed measurement.

The apparent fiber signature can then change when the background material changes, even though the textile itself is identical.

For recycling machines, belt composition should therefore be selected deliberately and included in calibration. A spectrally stable background that differs strongly from the textiles can improve segmentation and reduce ambiguity.

The same principle applies to multilayer garments: the layer beneath the visible surface can influence the measured signal if the outer fabric transmits significantly.

Oil and Finishing Chemicals Can Create Both Defects and Classification Interference

Textile finishing introduces oils, softeners, coatings, resins and functional treatments. Depending on chemistry, these materials may produce their own spectral signatures.

This can be useful if the objective is to verify whether a finish has been applied uniformly.

It can be problematic if the machine is trying to classify fiber composition and the finish dominates the measured spectrum.

The OEM should therefore characterize unprocessed fabric, correctly finished fabric and deliberately over- or under-treated samples separately.

A classifier should distinguish base fiber chemistry from expected process chemistry whenever both influence the image.

Kyptec Automation® KL-1412 for Controlled Fiber and Blend Analysis

For narrower material-analysis stations, the Kyptec Automation® KL-1412 25 MM SWIR Camera Lens provides an intermediate focal-length option within the same specialized SWIR family.

A 25 mm geometry can be useful where the system needs to examine a smaller textile region with stronger sensor utilization, such as blend verification, incoming-material identification, laboratory-to-production composition analysis or controlled contamination inspection.

The advantage is not simply magnification. The tighter FOV can give the classifier more pure pixels from a specific material region and fewer pixels from surrounding conveyor or adjacent fabric.

Fabric Speed Must Be Converted Into Motion During Exposure

Textile webs can move continuously at substantial line speed.

If a fabric moves at 2 m/s and camera exposure is 500 µs:

2000 mm/s × 0.0005 s = 1 mm movement during exposure.

If object-side sampling is 0.25 mm/pixel, that represents about four pixels of motion.

A broad composition measurement may tolerate some motion, but a narrow contamination streak or local moisture boundary can become blurred significantly.

Reducing exposure to 100 µs lowers movement to 0.2 mm.

The F1.4 maximum aperture available across the Kyptec Automation® SWIR Camera Lens portfolio provides useful light-collection margin when shorter exposure is necessary.

The illumination system still needs enough SWIR output to maintain classification signal at the required speed.

Moisture Mapping Should Measure Cross-Web Uniformity, Not Only Average Water Content

A valuable textile-drying metric is the difference between the wettest and driest sections of the web.

Suppose average residual moisture meets specification, but one edge consistently leaves the dryer wetter. The average value can hide that process imbalance.

Imaging makes it possible to evaluate:

cross-web mean moisture;

maximum local moisture;

minimum local moisture;

spatial standard deviation;

and persistent wet zones.

Research on wide textile webs specifically highlights the value of NIR hyperspectral imaging for monitoring spatial drying uniformity.

For process-control OEMs, this is a much more actionable output than one single moisture number.

Water Spots Can Be Detected Even on Visually Difficult Fabrics

Water spotting is difficult to inspect when fabric colour or pattern masks the visible difference.

Research on dyed fabrics found that near-infrared hyperspectral imaging could highlight moisture-related water spots across fabrics of different colours, with approximately 1459 nm emerging as a particularly useful band in that experiment.

This provides an important application lesson.

The best wavelength for a textile defect should be selected from the contrast between the defect and all acceptable fabric variation, not only from the defect itself.

If dye colour creates strong visible variability but limited SWIR variability in the chosen band, SWIR can simplify the classification problem substantially.

35 mm Geometry Can Support Secondary Textile Quality Control

After bulk sorting or web processing, a narrower quality-control station may need to inspect a selected section in greater detail.

The Kyptec Automation® KL-1414 35 MM SWIR Camera Lens provides a tighter-field option with 900–1700 nm, F1.4, 2 MP, 2/3-inch and C-Mount specifications. Textile is explicitly listed among its published applications.

This geometry can be particularly useful for controlled fiber-composition verification, localized finish inspection or small-region contamination analysis where maximum web width is no longer the main design constraint.

Kyptec Automation® therefore provides useful progression from wide inspection geometry to more concentrated material analysis within one SWIR optical family.

Textile Sorting Models Should Include Used and Aged Materials

Recycled textiles do not resemble pristine laboratory swatches.

Washing, wear, UV exposure, dirt, ageing, softeners and repeated drying can alter surface and material behaviour.

A model trained only on new fabric can therefore become unreliable when deployed on post-consumer garments.

The dataset should contain the actual material age and condition expected in production.

This is especially important for AI-based automated textile sorting because spectral class boundaries must remain stable across real-world variability, not merely between ideal reference samples.

Classification Should Include an Unknown Fiber Category

Textile recycling streams eventually contain materials that were not represented during training.

If an unfamiliar blend is forced into the nearest known class, it can contaminate the accepted recycling stream.

A stronger system includes an unknown or low-confidence condition. Material outside the validated spectral population can be diverted for secondary sorting or manual verification.

This is particularly useful when purity matters more than maximizing automatic recovery.

NIST's recent NIR textile reference work emphasizes the importance of representative, validated spectral datasets for automated textile feedstock identification, underscoring that robust classification depends heavily on the quality and diversity of the material library.

Purity and Recovery Are Better Metrics Than Overall Accuracy

A textile sorter with 97% overall accuracy may still perform poorly if the remaining errors systematically contaminate one valuable recycling stream.

For example, a cotton-rich output intended for cellulose recycling should be evaluated for residual polyester contamination, while a polyester stream should be evaluated for unwanted natural-fiber content.

Useful acceptance metrics include:

target-fiber recovery;

accepted-stream purity;

incorrect rejection of valuable material;

unknown-material rate;

and blend-class confusion.

These measures connect optical performance directly to recycling economics.

Why Kyptec Automation® Is a Strong SWIR Platform for Textile OEMs

The Kyptec Automation® SWIR Camera Lens collection provides five focal lengths—8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm—within a common 900–1700 nm, 2 MP, 2/3-inch, F1.4 and C-Mount architecture. Textile is specifically included among the published application areas for the Kyptec Automation® SWIR range.

For machine builders, this is useful because textile applications span very different optical geometries. A wide recycling conveyor can require the shorter focal lengths. A narrower fabric-composition station can use an intermediate lens. Detailed contamination or finishing inspection can justify 35 mm or 50 mm where greater stand-off or tighter sensor utilization is required.

Kyptec Automation® therefore offers a focused SWIR optical family that can support textile identification, process inspection and automated sorting without forcing every machine variant into the same field of view.

Frequently Asked Questions About SWIR Camera Lenses for Textile and Fiber Inspection

1. Can SWIR imaging distinguish cotton from polyester?

Yes. Cotton and polyester have different molecular compositions and therefore different NIR/SWIR spectral behaviour. Research has demonstrated clear differentiation between cotton and polyester and has also shown that hyperspectral imaging can estimate polyester concentration in blended textiles. Real production systems should nevertheless include multiple colours, constructions and aged samples during calibration.

2. Can SWIR identify natural and synthetic fibers automatically?

Potentially yes. Natural, regenerated and synthetic fibers can produce different spectral signatures because their chemical structures differ. Automated textile-sorting studies using approximately 900–1700 nm have demonstrated classification of several natural, synthetic and blended textile classes. The achievable number of classes depends on how spectrally distinct the materials are and how representative the training library is.

3. Can a SWIR system determine the percentage of polyester in a cotton-polyester blend?

Research indicates that NIR hyperspectral imaging can estimate polyester content across blended textiles, but this is a quantitative calibration problem rather than simple material identification. The model needs reference fabrics with known blend percentages covering the expected range. Published work has reported prediction errors of a few percentage points under controlled experimental conditions.

4. Why are low-percentage blend components harder to detect?

When one fiber represents only a small portion of the textile, the dominant material controls most of the measured spectral response. Spatial mixing and yarn construction can make the minor component even harder to isolate. If identifying a 5% or 10% blend component is commercially important, that exact concentration range must be included during system qualification.

5. Can SWIR textile sorting work on different fabric colours?

Often yes, because fiber classification can rely on infrared spectral features rather than visible colour alone. However, dyes can still influence parts of the infrared spectrum, so every major colour and dye family expected in production should be represented in training. The machine should prove that material classification remains stable when the visible colour changes.

6. Can SWIR detect moisture variation across a moving textile web?

Yes. NIR hyperspectral studies have demonstrated residual-moisture measurement across wide finished textile webs and have shown that imaging can reveal spatial drying non-uniformity. This makes SWIR useful not only for defect detection but also for process-control applications after washing, impregnation or drying.

7. Can SWIR detect water spots on dyed fabric?

Research has shown that water spots can become distinguishable in NIR hyperspectral imaging even on differently coloured fabrics. In one study, approximately 1459 nm provided particularly useful water-spot contrast while visible dye variation had less influence over the studied infrared range.

8. Does moisture interfere with fiber identification?

It can. Water has strong spectral behaviour in the SWIR region, so changing moisture content can alter a fabric's measured spectrum. If the goal is fiber sorting rather than moisture measurement, the classifier should be validated across the expected moisture range or the upstream process should standardize moisture before inspection.

9. Can SWIR identify textile contamination such as oil or foreign fibers?

Potentially, when the contaminant has a different SWIR response from the base textile. Oil, polymer contamination and foreign fibers should be treated as separate material classes because each produces different spectral behaviour. The smallest contamination region that must be rejected should also be defined before choosing the lens FOV.

10. Can SWIR identify fabric composition through printed patterns?

It can sometimes reduce the influence of visible printing because classification relies on infrared rather than visible colour, but printing inks and coatings can themselves modify SWIR response. Printed and unprinted examples of the same fabric should therefore be included during model development rather than assuming the print is spectrally invisible.

11. Which SWIR focal length is suitable for a wide textile sorting conveyor?

Broad conveyors generally favor shorter focal lengths when camera height is limited. The Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens provides the widest focal-length geometry in the current Kyptec Automation® SWIR portfolio, while 12.5 mm can provide stronger object sampling where slightly less coverage is needed. Selection should be based on actual belt width, working distance and smallest material region.

12. When is a 25 mm SWIR lens better for fiber inspection?

The Kyptec Automation® KL-1412 25 MM SWIR Camera Lens becomes useful when the inspection width is smaller and greater sensor utilization over individual textile regions is desirable. It can suit blend verification, incoming material inspection and controlled composition-analysis stations better than a very wide conveyor configuration.

13. Can thin fabric be difficult to classify with SWIR?

Yes. Thin textiles can allow the background or underlying layer to contribute to the measured signal, creating a mixed spectrum. Very thin fabrics and small blend fractions have also been reported as more challenging in textile-sorting research. The conveyor background and material thickness should therefore be included during validation.

14. Why can a folded fabric produce a different SWIR response from the same flat fabric?

Folding changes surface orientation, scattering, local thickness and shadowing. A classifier trained only on perfectly flat fabric can therefore become sensitive to geometry rather than chemistry. Recycling systems should include wrinkled, folded and worn garments in the training and validation library.

15. How does textile web speed affect SWIR inspection?

Higher line speed reduces the available exposure time before motion blur becomes significant. The amount of fabric movement during exposure should be compared with the smallest defect or contamination width. The F1.4 aperture available across the Kyptec Automation® SWIR family provides useful optical throughput for shorter exposures, but illumination strength must also be designed accordingly.

16. Can SWIR help control textile drying rather than just sort fabrics?

Yes. Imaging can map residual moisture across the entire inspected area and identify cross-web drying non-uniformity. This can support feedback to drying or finishing processes rather than merely generating a reject decision. Controlled studies on wide textile webs have demonstrated this type of moisture mapping.

17. Should an automated textile sorter force every garment into a known fiber class?

No. A low-confidence or unknown category is valuable because real waste streams contain unusual blends, coatings, multilayer garments and materials absent from the training database. Diverting uncertain items for secondary sorting can protect the purity of high-value recycling streams.

18. How should an OEM validate a SWIR textile-sorting machine?

Use independently identified samples covering the major pure fibers, required blend ratios, colours, dyes, finishes, garment ages, thicknesses and moisture conditions expected in production. Test folded and flat material, edge positions and maximum conveyor speed. Report accepted-stream purity and recovery for each important fiber class rather than relying only on overall classification accuracy.

19. What information should be provided when selecting a SWIR camera lens for textile inspection?

Provide fabric or conveyor width, SWIR camera sensor dimensions, available working distance, smallest contamination or composition region, web speed, expected material classes, fabric thickness range, moisture conditions and whether the machine is performing fiber identification, blend analysis, contamination inspection or moisture mapping. These inputs allow focal length to be selected from the actual textile process rather than from the application name alone.

20. Why is Kyptec Automation® a strong option for textile and fiber SWIR inspection?

Kyptec Automation® provides a dedicated SWIR Camera Lens collection spanning 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm focal lengths while maintaining 900–1700 nm operation, 2 MP resolution class, 2/3-inch format, F1.4 and C-Mount across the current portfolio. Textile is explicitly included among the published applications. This gives OEMs a coherent optical platform for wide textile sorting, intermediate material-identification stations and tighter contamination or process-quality inspection without moving outside the specialized Kyptec Automation® SWIR Camera Lens family.

Conclusion

A SWIR camera lens for textile and fiber inspection becomes valuable when the quality decision depends on material composition, water content or contamination rather than only on what the fabric looks like. Natural and synthetic fibers have different molecular structures, and those differences can create useful spectral signatures across the NIR/SWIR region. Research has demonstrated differentiation between cotton and polyester, estimation of polyester concentration in blended fabrics, classification of multiple textile types for recycling and real-time sorting across approximately 900–1700 nm.

The same spectral sensitivity can support textile-process control. Residual moisture can be mapped across a wide fabric web, localized water spots can be distinguished on coloured textiles, and drying non-uniformity can be visualized spatially rather than reduced to a single average moisture reading. Contamination, finishing chemicals and foreign fibers can also become measurable when their SWIR response differs sufficiently from the underlying textile.

For a machine builder, the main challenge is converting those spectral differences into reliable production decisions. Fabric folds, thickness, dyes, coatings, moisture, ageing and conveyor background all influence the measured signal. Wide FOV increases throughput but reduces the pixels available for small material regions. High web speed reduces allowable exposure. Thin textiles can introduce background mixing, and low-percentage blends can become difficult to separate from the dominant fiber. These variables should be incorporated into the optical specification and training database rather than treated as exceptions after installation.

The Kyptec Automation® SWIR Camera Lens collection provides a particularly useful foundation for this process because its five focal lengths allow the same dedicated SWIR category to address several textile geometries. The Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens can support broad inspection coverage where large web width is essential. The Kyptec Automation® KL-1410 12.5 MM SWIR Camera Lens provides a useful compromise between width and material sampling. The Kyptec Automation® KL-1412 25 MM SWIR Camera Lens can serve more controlled fiber-composition analysis, while longer focal lengths can support tighter quality-control stations. The common 900–1700 nm, F1.4, 2 MP, 2/3-inch and C-Mount architecture gives OEMs a consistent specialized optical platform for building several types of textile inspection machine.

For textile manufacturers, recycling-equipment builders and automation OEMs, the strongest engineering principle is to define the material decision first and the lens geometry second. Establish whether the machine must identify pure fiber, quantify a blend, map residual moisture, detect contamination or sort mixed garments; determine the smallest region and lowest material percentage that changes the production decision; then design wavelength selection, FOV, working distance, exposure and classification around that requirement. When those elements are engineered as one system, Kyptec Automation® SWIR Camera Lenses can provide a strong optical foundation for textile material identification, moisture monitoring, contamination inspection and automated fabric sorting beyond the limitations of conventional visible imaging.