SWIR Camera Lens for Plastic Identification and Recycling: Polymer Sorting Beyond Visible Color

Plastic recycling becomes far more difficult when materials that look similar must be separated according to polymer chemistry rather than visible appearance. A clear bottle, translucent film, colored container and molded component can all belong to different polymer families, while plastics made from the same resin may appear in many colors and surface finishes. Conventional color imaging can identify shape, color, labels and visible contamination, but it cannot reliably determine whether an object is polyethylene, polypropylene, polyethylene terephthalate or another polymer simply from RGB appearance. Short-wave infrared imaging offers a different route because polymers interact with infrared wavelengths according to their molecular structure, producing spectral absorption and reflectance patterns that can be used for material identification. Studies using approximately 900–1700 nm hyperspectral imaging have demonstrated discrimination among common post-consumer plastics and have shown that spectral information can separate materials that visible appearance alone cannot classify reliably.

For engineers and recycling-equipment buyers searching for a SWIR camera lens for plastic sorting, polymer identification camera lens, SWIR lens for recycling machines, plastic sorting machine vision lens, 900–1700 nm lens for material identification, or infrared lens for PE, PP and PET sorting, the lens should be treated as part of a material-classification system rather than a simple viewing component. The Kyptec Automation® SWIR Camera Lens collection provides 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm focal lengths within a focused 900–1700 nm, 2 MP, 2/3-inch, F1.4 and C-Mount portfolio. That range allows machine builders to adapt optical geometry to broad conveyor belts, localized sorting zones and laboratory-to-production development while remaining within a dedicated SWIR lens family.

Why Visible Color Is a Weak Basis for Polymer Identification

Visible imaging primarily records how an object reflects wavelengths perceived as red, green and blue. Two plastics made from different polymers can share nearly identical visible color because pigments dominate their appearance. Conversely, the same polymer can be manufactured as clear, white, blue, green or many other colors.

A recycling line based only on visible color can therefore sort objects by appearance without actually identifying their resin chemistry. This is a serious limitation when recycled material purity matters, because visually similar polymer streams may behave differently during melting, extrusion and remanufacturing.

SWIR imaging shifts the decision from visible appearance toward material-dependent spectral behavior. Molecular bonds within different polymers absorb and reflect infrared radiation differently, creating wavelength-dependent patterns that can be used as spectral fingerprints. Research using 900–1700 nm imaging has successfully classified several recyclable plastics including PP, PET, PS, HDPE, LDPE and PVC.

The important point for an industrial buyer is that SWIR does not identify plastic because one resin is simply “brighter” than another. It works because the shape of the spectral response across multiple wavelengths can differ.

Polymer Sorting Depends on Spectral Shape, Not One Universal Brightness Level

A common misunderstanding is to assume that one polymer always appears dark while another always appears bright in SWIR. In practice, polymer identification usually relies on several wavelength-dependent features rather than one absolute intensity threshold.

A piece of PET may reflect strongly at one wavelength and absorb more strongly at another. PP can produce a different curve, while PE may show related but distinguishable spectral behavior. By measuring these patterns, classification software can distinguish materials even when overall brightness changes because of surface geometry or illumination.

This is why hyperspectral and multispectral plastic sorting systems evaluate spectral signatures rather than relying only on a single grayscale image. Research has demonstrated that PET and PVC can be separated using information derived from the 900–1700 nm region, while post-consumer thermoplastic studies have classified multiple recyclable resin types using spectral differences in the same general wavelength interval.

The SWIR camera lens must therefore preserve the spatially resolved signal from each plastic piece so the sensor receives enough spectral information for classification.

Why PE and PP Can Be Difficult but Important to Separate

Polyethylene and polypropylene are both polyolefins and can be visually similar. Depending on the product form, color and contamination, they may be difficult to distinguish using conventional machine vision. Their separation matters because the purity of recycled polymer streams influences the consistency and value of the resulting secondary material.

Spectral imaging has been investigated specifically for distinguishing polyolefin waste streams, including PE and PP, and research has shown that SWIR/NIR information can support quality-control classification in recycling processes.

For machine builders, this means the optical system must do more than detect the presence of a plastic item. It needs enough spectral signal and spatial sampling to classify that item according to resin family.

If a conveyor contains small flakes rather than complete bottles or containers, each piece must occupy enough pixels for the classification algorithm to obtain useful interior measurements without too much mixing from the belt or neighboring fragments.

Why PET Sorting Is a Major SWIR Recycling Use Case

PET recycling depends heavily on stream purity. Contamination with incompatible polymers can degrade the quality of recovered material, particularly when flakes are being processed for demanding reuse applications.

SWIR hyperspectral imaging has been studied specifically for PET recycling and for separating PET from other polymer contaminants. Research has shown that spectral information can support identification of PET through different phases of the recycling chain and can be used to improve quality control of recovered PET flakes.

For a production machine, the optical requirement depends on whether the system sees whole packages, shredded flakes or another material form. Whole objects may need a large conveyor field, whereas flakes demand higher spatial sampling because each item is small.

This distinction affects focal-length selection even when the underlying polymer classification wavelengths remain similar.

Why a SWIR Camera Lens Is Critical to Spectral Plastic Identification

The classification algorithm can only work with information delivered to the camera. If the optical system produces excessive blur, poor edge performance, insufficient signal or inadequate field coverage, the spectral classifier receives compromised data.

A SWIR lens for plastic identification therefore has several jobs simultaneously. It must support the operating wavelength region, match the camera format, provide enough light for the required exposure time, and map each plastic piece onto sufficient sensor area for reliable analysis.

The current Kyptec Automation® SWIR portfolio is specified for 900–1700 nm operation, 2 MP resolution, 2/3-inch format, F1.4 aperture and C-Mount. These parameters provide a useful basis for compatible material-identification systems operating in the same spectral region widely used in polymer-classification research.

Broad Conveyor Sorting Requires a Different Lens Geometry Than Flake Inspection

A recycling machine handling bottles and large objects may need to cover a wide conveyor so every item can be located and classified before reaching the ejection zone. A flake-sorting machine may inspect much smaller pieces and therefore needs greater image scale.

For broad conveyor views, the Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens provides the shortest focal length in the current portfolio. Its verified specifications include 900–1700 nm, 2 MP, 2/3-inch format, F1.4 and C-Mount.

The engineering trade-off is that a wider field spreads the available pixels across more conveyor area. That may be acceptable for large containers but insufficient for small flakes. Focal length should therefore be chosen from the smallest plastic piece that must be classified reliably, not just from the total belt width.

Why Background Selection Matters in Plastic Sorting

The conveyor belt or chute becomes part of every spectral measurement. If a plastic piece occupies only a few pixels, edge pixels can contain a mixture of plastic and background signal. This spectral mixing can distort the polymer signature and reduce classification confidence.

A suitable background should therefore provide stable contrast against the materials being inspected while avoiding spectral characteristics that interfere with the classifier.

The optical geometry should also ensure that small pieces occupy enough pixels to provide interior regions dominated by the plastic itself rather than the belt beneath it.

This is one reason why image scale is so important in material classification. Spatial resolution and spectral purity are connected: insufficient magnification increases the percentage of mixed pixels around every small object.

Using 12.5 mm Where Wide Coverage Still Needs Better Plastic Representation

A conveyor may require substantial field coverage without needing the maximum width provided by the shortest focal length. In that case, a slightly tighter view can devote more sensor samples to each plastic item.

The Kyptec Automation® KL-1410 12.5 MM SWIR Camera Lens provides an intermediate wide-field option in the Kyptec Automation® SWIR family. Its live product page confirms the same F1.4, 2/3-inch and C-Mount platform used across the current range.

This type of geometry can be useful where bottles, containers or larger fragments need classification across a broad sorting area but excessive background should be avoided.

Color Independence Does Not Mean Surface Condition Is Irrelevant

SWIR can reduce dependence on visible color, but it does not make the system immune to physical surface variation. Dirt, labels, liquids, coatings, scratches, oxidation and residue can modify the measured spectrum.

A clean laboratory sample may therefore classify more easily than post-consumer waste.

Production training data should deliberately include real-world variation: clear plastic, colored plastic, weathered material, partially dirty surfaces, crumpled packaging and multiple suppliers. The classifier must learn the polymer features that remain stable across those variations rather than memorizing ideal samples.

Recent research into post-consumer plastic classification emphasizes precisely this challenge: real recycling streams contain varied household materials rather than uniform virgin polymer specimens.

Labels and Sleeves Can Hide the Polymer Beneath Them

A bottle may be PET while its label is made from a different material. If the SWIR system images the label rather than the exposed bottle surface, the classifier may identify the covering material instead of the underlying container.

This creates a segmentation problem as much as a spectroscopy problem.

The system should collect enough spatial information to identify exposed regions of the principal object and avoid basing the entire decision on a label, cap or attached component. Multiple pixels or regions can be evaluated so the classification represents the dominant material rather than one localized accessory.

For whole-object recycling, this can improve robustness considerably compared with using a single spectral point.

Moisture Can Alter the Spectrum of Recycled Plastic

Post-consumer plastic may arrive wet after washing, storage or exposure to weather. Water has strong absorption features in SWIR, so surface moisture can modify the measured spectrum and potentially interfere with polymer classification.

A classifier trained only on dry samples may therefore experience reduced confidence on wet material.

This does not mean wet plastic cannot be sorted, but moisture variation should be included during development. Dry and wet versions of representative polymers should be captured so the classification model can distinguish water-related spectral changes from the underlying resin signature.

This is particularly important when washing occurs immediately before optical sorting.

Why Transparent Plastic Can Still Have a SWIR Spectral Signature

A transparent bottle may transmit much of the visible spectrum and appear almost colorless, but transparency to visible light does not mean the polymer has no infrared absorption characteristics.

Spectral imaging can reveal molecular absorption features that are not obvious in visible appearance. This is one reason transparent PET and other clear polymers can still be candidates for spectral classification.

Research using hyperspectral imaging has classified transparent and colored plastic bottles using information extending through the near-infrared/SWIR region.

For transparent objects, however, background and geometry become particularly important because radiation can interact with both the plastic and whatever lies behind it.

Black Plastic Is an Important Limitation That Must Be Understood

SWIR/NIR plastic sorting is powerful, but it is not universally effective on every dark plastic. Carbon-black pigments can strongly absorb near-infrared radiation and suppress the underlying polymer spectral signature, making conventional NIR/SWIR reflectance classification difficult.

This limitation is well documented in recycling research and is one reason black-plastic identification is treated as a specialized problem rather than a straightforward extension of ordinary polymer sorting.

A technically responsible SWIR system design should therefore not promise reliable identification of every black polymer without sample validation. Dark plastics using pigments that do not suppress the relevant spectral features may behave differently, but carbon-black-containing materials require careful testing.

For Kyptec Automation® buyers, the practical rule is simple: test the actual black plastic waste stream before freezing the optical architecture.

Why Spectral Libraries Need Real Production Samples

A polymer classifier needs reference data describing what each material looks like across the selected wavelengths. Building that library from pristine virgin plastics alone can create poor generalization to real waste.

Post-consumer plastics may be scratched, faded, chemically aged, dirty, printed, stretched or partially degraded. Their spectral characteristics can shift while remaining recognizably associated with the same polymer family.

A robust dataset should therefore contain many samples from different product types, manufacturers, colors and usage histories.

This is especially important when separating chemically similar materials such as HDPE and LDPE or distinguishing contamination within a high-purity stream.

Hyperspectral Development Can Lead to Simpler Multispectral Production Systems

During research and development, a hyperspectral camera can measure many narrow wavelength bands and help identify which regions carry the most useful polymer separation. Once those wavelengths are known, a production machine may not always need every band.

Feature-selection research has demonstrated that selected subsets of wavelengths can retain useful plastic-classification information.

This creates a practical development path: first characterize polymers broadly, identify the strongest discriminating bands, then simplify the production architecture where appropriate.

The SWIR camera lens remains central because whichever wavelengths are chosen must still pass through the optical system with sufficient signal and image quality.

Using 25 mm for Smaller Plastic Parts and Controlled Sorting Zones

When the inspection area is narrower or plastic pieces are relatively small, a medium focal length can devote more sensor area to each item.

The Kyptec Automation® KL-1412 25 MM SWIR Camera Lens can be evaluated for controlled plastic-classification stations where the full conveyor width does not need to occupy one extremely wide frame.

A tighter field can improve the number of pixels available per fragment, helping the algorithm isolate interior polymer pixels and reduce background mixing.

This can be particularly useful in secondary quality-control stages where a primary sorter has already narrowed the material stream and the next station verifies purity more closely.

Why Sorting Speed Changes the Optical Requirement

Industrial recycling is a high-throughput environment. Objects move continuously, and the vision system has limited time to acquire, classify and send the result to an ejection mechanism.

As exposure time becomes shorter, fewer photons reach the sensor. Signal-to-noise ratio can fall, especially if the selected spectral band has relatively low illumination or strong material absorption.

The F1.4 aperture available across the Kyptec Automation® SWIR Camera Lens portfolio provides useful light-gathering capability for compatible systems, but illumination power, object velocity, sensor sensitivity and aperture still need to be balanced together.

The production test should always be performed at actual conveyor speed. A classifier that works on stationary plastics may fail once motion blur, vibration and shortened exposure reduce spectral or spatial quality.

The Classifier Must Know When It Does Not Recognize a Material

A real recycling stream contains more than the polymer classes used during training. Paper, metal, wood, glass, textiles, multilayer packaging, unusual resins and unknown contaminants may enter the field.

A classification system that is forced to assign every unknown object to PE, PP or PET can contaminate recovered material streams.

Strong production architectures therefore include confidence thresholds or rejection logic so uncertain spectra can be routed to a residual stream rather than given a false high-confidence label. Recent SWIR plastic-sorting research has specifically explored out-of-distribution detection for this reason.

This is an important buyer consideration because purity depends not only on recognizing known plastics, but also on avoiding confident misclassification of unknown materials.

Sorting Purity Is Often More Important Than Raw Classification Accuracy

A system can report impressive overall accuracy while still allowing enough contamination into one output stream to reduce its recycling value.

For example, if the objective is producing high-purity PET flakes, the most important metric may be how much non-PET material remains after sorting rather than average recognition across every class.

Validation should therefore measure precision, recall, contamination rate and recovered yield for each commercially important polymer stream.

Optical design supports this by ensuring small contaminants are large enough in the image to classify and by preserving sufficient spectral signal across the full belt width.

Why 35 mm and 50 mm Can Be Useful for Secondary Purity Inspection

After a broad primary sorting stage, a secondary station may inspect a narrower material stream for residual contamination. Here, tighter optical framing can be more valuable than maximum conveyor coverage.

The Kyptec Automation® KL-1414 35 MM SWIR Camera Lens can support a comparatively narrow inspection field where individual pieces need stronger representation. The Kyptec Automation® KL-1416 50 MM SWIR Camera Lens extends this approach where greater stand-off or an even tighter region is required.

These longer focal lengths should not be viewed as inherently better for polymer identification. Their value comes from geometry: they can allocate more of the available sensor area to smaller material regions when the machine layout permits.

Why Kyptec Automation® Is a Strong Optical Platform for Plastic Sorting Development

Plastic sorting machines vary widely in conveyor width, fragment size, camera distance and required throughput. A whole-container sorting line may need broad coverage, while a flake-purity system may require a much tighter field.

The Kyptec Automation® SWIR Camera Lens collection gives OEM machine builders five focal lengths—8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm—within a dedicated 900–1700 nm SWIR platform. The live portfolio also identifies material identification among the applications for these lenses.

That combination is particularly useful for recycling system development because engineers can keep the optical category aligned with the required SWIR spectral region while changing focal length according to the physical sorting stage. Kyptec Automation® therefore provides a practical lens family for moving from laboratory polymer characterization to wider primary sorting or more tightly framed secondary purity inspection.

Frequently Asked Questions About SWIR Camera Lenses for Plastic Identification and Recycling

1. Can SWIR identify different types of plastic?

Yes. Many polymers exhibit distinct wavelength-dependent absorption and reflectance patterns in the near-infrared and SWIR regions. These spectral signatures can be used to distinguish materials such as PET, PP, PE, PS and PVC under suitable imaging conditions. Research using approximately 900–1700 nm hyperspectral data has demonstrated classification of multiple common recyclable polymers.

2. Why is SWIR better than a normal color camera for polymer sorting?

A visible camera primarily distinguishes color, shape and surface appearance, while SWIR imaging can measure material-dependent spectral behavior. Two polymers that have the same visible color can still produce different SWIR signatures because their molecular structures interact differently with infrared wavelengths. SWIR therefore adds chemical-composition information that RGB imaging cannot reliably provide by itself.

3. Can SWIR separate polyethylene from polypropylene?

SWIR/NIR spectral imaging has been studied specifically for polyolefin sorting, including separation and quality assessment of PE and PP streams. Because the two materials are chemically related, robust classification requires representative spectral data and careful calibration rather than one simple brightness threshold.

4. Can a SWIR camera distinguish PET from PVC?

Yes, under suitable conditions. Research using approximately 900–1700 nm reflectance spectra demonstrated that PET and PVC can be distinguished using selected spectral information. For production equipment, the classifier should still be validated against real post-consumer material, including labels, dirt, color variation and different product histories.

5. Does plastic color affect SWIR identification?

Visible color generally matters less to spectral polymer identification than it does to RGB classification, but pigments and additives can still influence SWIR response. Clear, colored and weathered examples of each polymer should therefore be included during training and validation. The objective is to learn stable resin-related spectral features rather than color-specific appearance.

6. Can SWIR identify transparent plastic bottles?

Yes, transparent plastics can still have characteristic infrared absorption features even though they transmit much of the visible spectrum. The system must account for background contribution because radiation may pass through the object and interact with the belt or material behind it. Suitable segmentation and optical geometry are therefore especially important for clear containers.

7. Why are black plastics difficult to sort with SWIR?

Many black plastics contain carbon black, which strongly absorbs near-infrared radiation and can suppress the spectral signature of the underlying polymer. This makes standard SWIR/NIR reflectance classification difficult. Actual dark-plastic samples should therefore be tested rather than assuming that every black polymer can be identified reliably in 900–1700 nm imaging.

8. Can wet plastic still be identified with a SWIR sorting system?

It can be possible, but water itself has strong SWIR absorption features and can alter the measured polymer spectrum. A classifier intended for washed or outdoor waste should therefore include realistic wet samples during calibration. Otherwise, the system may interpret moisture variation as a polymer difference or lose confidence compared with dry training data.

9. What focal length is best for a plastic sorting machine?

There is no universal focal length. The correct choice depends on conveyor width, working distance and the smallest plastic piece that must be classified. Shorter focal lengths such as the Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens are useful for broad fields, while 25 mm, 35 mm or 50 mm options can provide tighter framing where small flakes or secondary purity inspection require more sensor pixels per object.

10. How many pixels should cover a plastic flake for reliable classification?

There is no fixed number because performance depends on spectral contrast, sensor noise and classifier design. However, a flake should ideally occupy enough pixels that several measurements come from its interior rather than edges mixed with the conveyor background. Very small objects represented by only a few pixels are more vulnerable to spectral mixing and misclassification.

11. Can one SWIR wavelength identify every type of plastic?

Usually not reliably. Polymer identification generally benefits from comparing responses at multiple wavelengths because spectral shape contains more material information than one intensity measurement. During development, hyperspectral data can help identify the wavelengths that best separate the required polymer classes, after which a simpler multispectral architecture may sometimes be possible.

12. Can SWIR identify plastic through a label or sleeve?

The system primarily measures the material visible to the optical path. If a label or sleeve covers the container, its spectrum can dominate the measurement. Whole-object classification should therefore evaluate multiple regions and identify exposed areas of the main polymer where possible rather than relying on one covered pixel or point.

13. Does a SWIR plastic sorter need hyperspectral imaging?

Not always. Hyperspectral imaging is valuable during development because it captures many wavelength bands and helps characterize polymer signatures. Once the most discriminating wavelengths are known, some production systems may use fewer selected bands. The required approach depends on how many polymers must be separated and how much real-world variation the sorter encounters.

14. What should I test before selecting a SWIR camera lens for recycling equipment?

Define conveyor width, working distance, smallest material size, camera format, line speed, target polymer classes, expected contamination and illumination strategy. Then test representative plastics across colors, ages, labels, moisture conditions and surface contamination. For compatible systems, Kyptec Automation® offers 8.5 mm through 50 mm SWIR Camera Lens options so focal length can be matched to actual sorting geometry after these requirements are established.

15. How should a plastic sorting system handle unknown materials?

The classifier should not be forced to assign every spectrum to a known polymer. Confidence thresholds, reject classes or out-of-distribution detection can route uncertain items away from high-purity output streams. This is especially important in real recycling facilities because unusual plastics and non-plastic contaminants are unavoidable. The goal is not only maximum recognition rate but protection of final stream purity.

Conclusion

The real value of SWIR in plastic recycling is that it shifts sorting from visible appearance toward material identity. PE, PP, PET and other polymers can look almost identical in ordinary images while producing different spectral behavior in the approximately 900–1700 nm region. That difference creates the foundation for automated polymer classification, recycled-stream quality control and higher-purity material recovery. Research across post-consumer plastics confirms that SWIR/NIR hyperspectral information can distinguish several common resin classes and can support practical recycling applications.

Reliable production performance still requires more than a spectral classifier. The camera lens must support the required SWIR wavelengths, provide enough signal at production exposure times, match the sensor, cover the required sorting width and allocate sufficient pixels to the smallest fragment. Wet material, labels, background mixing, surface contamination and carbon-black plastics must also be included in validation because real recycling streams are far less controlled than laboratory samples.

The Kyptec Automation® SWIR Camera Lens collection provides a focused optical family with 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm focal lengths specified around 900–1700 nm, 2 MP resolution, 2/3-inch format, F1.4 aperture and C-Mount. This range gives recycling-equipment OEMs the flexibility to design broad primary sorting, intermediate material-identification stations and tightly framed secondary purity inspection while remaining within one dedicated SWIR optical platform.

For buyers developing plastic identification systems, the most effective approach is therefore to characterize the actual polymer stream first, determine which wavelengths produce reliable material separation, define the smallest object that must be classified, then choose the Kyptec Automation® SWIR Camera Lens focal length that converts those spectral differences into sufficiently large, stable and repeatable image regions on the sensor. When polymer physics, spectral classification and optical geometry are designed together, SWIR becomes a practical tool for sorting plastics by what they are—not simply by what color they appear.