LED de 1000-1750NM de giro
Industrial plastic sorting systems must distinguish between materials that can appear remarkably similar when inspected using ordinary visible-light cameras. In recycling operations, polymer processing facilities, automated material separation equipment and manufacturing quality control systems, the appearance of a plastic component does not necessarily reveal its chemical composition. Two plastic fragments may have nearly identical colours, surface textures and dimensions while belonging to different polymer families, and products manufactured from the same polymer can appear completely different because of pigments, additives, labels and surface treatments. This creates an important purchasing decision for manufacturers developing optical sorting machinery: whether a conventional visible machine vision camera provides enough information for the intended separation task or whether short wave infrared imaging is necessary to identify material differences that ordinary colour and intensity measurements cannot reliably distinguish.
The comparison between visible machine vision and SWIR imaging should not be reduced to a general claim that infrared systems are more advanced or inherently more accurate. Each imaging technology measures different optical characteristics, and its suitability depends on the physical information required for the sorting decision. A visible camera can identify object boundaries, distinguish colours, recognize packaging shapes and locate printed labels with useful spatial detail. These capabilities remain valuable when the manufacturing objective involves appearance, product type, object positioning or separation according to an established visual characteristic. No entanto, when the objective is to distinguish polyethylene from polypropylene, separate polyethylene terephthalate from another visually similar material or identify polymer contamination within a mixed recycling stream, visible appearance may not provide dependable evidence of material identity.
SWIR imaging provides another source of information by examining how plastic materials respond to wavelengths beyond the visible spectrum. Many polymers exhibit wavelength-dependent absorption and reflection characteristics associated with their molecular structure. These characteristics may provide useful spectral differences even when the same materials cannot be distinguished consistently through red, green and blue camera channels. Yet a material that is identifiable under laboratory SWIR illumination does not automatically become easy to classify on a high-speed production conveyor. Surface contamination, specimen thickness, pigment composition, background contribution, illumination variation and spectral measurement quality can all influence the acquired information.
For industrial equipment buyers, the most important question is therefore not whether SWIR is categorically superior to visible imaging. It is whether the additional spectral information produces a measurable improvement in the particular sorting result that matters commercially. A recycling facility separating clear bottles from coloured bottles has a different requirement from a plant attempting to remove polypropylene contamination from a polyethylene stream. One facility may need to identify complete packages before mechanical separation, while another must classify small shredded fragments with irregular shapes and limited usable surface area. The correct optical architecture depends on these different physical tasks and should be established through representative material trials before substantial equipment investment.
The lens is an essential part of that measurement chain. Suitable machine vision lenses allow compatible industrial cameras to capture the object field and preserve the image information needed for segmentation or material classification. In visible inspection, optical design concentrates on the wavelengths used to observe surface appearance, while SWIR inspection requires camera-side optics suitable for the intended infrared bands. A conventional lens should not be assumed to deliver adequate spectral transmission or image quality at 900 para 1700 nm merely because its mount fits the infrared camera. The selected optical configuration must also provide the required coverage, focus and sampling of individual plastic pieces.
Kyptec Automation® offers a dedicated family of SWIR camera lenses within its Machine Vision Lens portfolio, with focal lengths of 8.5 milímetros, 12.5 milímetros, 25 milímetros, 35 mm and 50 milímetros. The published SWIR models are specified for the 900 para 1700 faixa nm, 2 MP resolution class, 2/3-inch sensor format, F1.4 aperture and C-mount interface. These focal-length choices provide a practical basis for designing compatible infrared imaging stations with different conveyor widths, working distances and object dimensions. They also allow OEM manufacturers to identify appropriate camera-side optical specifications before determining whether a particular spectral classification process meets the required purity and throughput targets.
The purpose of this article is to compare visible and SWIR machine vision from the perspective of real industrial sorting decisions rather than repeat a general description of polymer infrared absorption. It examines how to compare both imaging methods using the same material population, determine which measurements contribute genuine information, identify when visible and infrared observations are complementary and establish whether a more complex system produces a commercially meaningful improvement. It also explains how wavelength selection, lens geometry, spatial sampling, classification uncertainty and rejection timing should influence equipment procurement. The objective is to develop a sorting architecture that separates the required materials consistently under production conditions, without assuming that additional cameras or spectral bands automatically guarantee better recycling performance.
Comparing Visible and SWIR Material Contrast Under Real Sorting Conditions
A meaningful comparison begins by separating object recognition from material identification. Visible machine vision can identify information about what an object looks like, including its colour, apparent shape, printed markings, surface texture and position on a conveyor. These features can be highly informative when the sorting process deals with recognizable packaging formats or products with established visual distinctions. Por exemplo, a production line receiving a restricted range of known plastic containers may use visible appearance to distinguish product categories reliably, provided the relationship between appearance and material identity is controlled. The same approach may become unreliable when the feed contains objects from many suppliers whose packaging designs, pigments and polymer compositions vary independently.
The critical weakness of appearance-based polymer sorting is not that visible cameras produce poor images. A camera may deliver excellent resolution, stable colour response and highly accurate object boundaries while the underlying resin remains ambiguous. Pigments can make two different polymers appear identical, while products manufactured from one polymer can exhibit many visible colours. A classifier trained on these correlations may perform impressively on familiar products but fail when new packaging designs enter the recycling stream. The resulting error is a limitation of the information available to the model rather than necessarily a deficiency in its artificial intelligence architecture.
SWIR imaging changes the information available for classification because the measured response can include polymer-dependent spectral characteristics. Different resin families can exhibit distinguishable absorption features across portions of the near-infrared and SWIR regions, allowing material classification to rely less heavily on ordinary colour. No entanto, the underlying spectral response is not an absolute material label. Its measured appearance depends on wavelength, caminho óptico, sample thickness, illumination and the camera system. A material may be distinguishable using several carefully selected wavelengths but remain ambiguous when observed through a single broadband infrared image.
This distinction is fundamental for equipment manufacturers. A broadband SWIR image may reveal visual contrast unavailable to an RGB camera, but it does not necessarily provide enough information to separate all required polymer classes. Reliable material classification may require multiple spectral measurements, either through a suitable multispectral arrangement or through hyperspectral acquisition. The engineering requirement is to establish which wavelength-dependent differences matter for the particular material pairs rather than assume that an infrared camera automatically identifies a polymer.
For OEM manufacturers evaluating SWIR camera lenses, Kyptec Automation® KL-1414 provides a 25 mm focal length, 2 MP resolution class, 2/3-inch image format, F1.4 aperture and C-mount interface, with a published 900 para 1700 nm operating range. This specification provides a useful camera-side optical option for compatible infrared equipment observing a localized material region. The achievable classification performance must still be established using the intended camera, illumination wavelengths and representative plastic samples, because the lens provides image formation rather than identifying the polymer by itself.
A proper technology comparison should examine the same physical materials through both imaging methods. If visible images are collected from one set of clean laboratory samples while SWIR images are collected from another set containing more diverse materials, the resulting classification differences cannot be attributed confidently to spectral capability alone. The same representative pieces should be identified independently, imaged under controlled conditions and evaluated using processing methods appropriate to each modality. This provides a stronger basis for deciding whether SWIR contributes information unavailable through the visible system.
The material population should reflect the actual recycling or manufacturing process. A stream containing clean, rigid, unlabelled fragments presents a different imaging problem from one containing crushed containers, transparent films, printed packaging, food residue and mixed pigments. The sorting technology should be evaluated against those conditions rather than against idealized examples selected because they produce attractive infrared contrast. A classifier may achieve high accuracy on isolated samples while failing to maintain acceptable purity when multiple objects overlap or when a label conceals the underlying material.
The comparison should also distinguish classification accuracy from the final sorting result. A camera may classify the visible portion of an object correctly but still provide an insufficient estimate of its physical position for downstream separation. Another camera may obtain a useful material spectrum but fail to associate that spectrum with the correct fragment on a crowded conveyor. In both cases, the image classification can appear technically successful while the machine produces an incorrect separation decision.
This is why visible and SWIR systems should be compared according to actual sorting outcomes. The relevant questions include how many required materials are recovered, how much unwanted material enters the accepted stream, which material classes are most frequently confused and whether the equipment maintains those results as feed composition changes. These measurements describe commercial performance more meaningfully than the visual quality of individual demonstration images.
A further difference concerns spatial information. Visible cameras can provide high-detail images of object contours, printed features and surface boundaries, while a particular SWIR system may provide different spatial resolution depending on its sensor and spectral architecture. If a spectral measurement includes pixels containing both a plastic fragment and the conveyor background, the recorded response can become a mixture rather than a clean representation of one material. This is particularly important for small flakes and thin fragments whose dimensions approach the effective spatial sampling of the optical system.
For sorting equipment designers evaluating industrial SWIR lenses, Kyptec Automation® KL-1412 provides a 12.5 mm focal-length SWIR configuration with the published 900 para 1700 nm operating range and compatibility specifications for a 2/3-inch, 2 MP, C-mount imaging arrangement. A shorter focal length can provide broader coverage at a given working distance, but the resulting field must still allocate sufficient image samples to the smallest plastic pieces that require classification. When the objective involves very small fragments, a narrower field, different camera arrangement or additional sensing positions may be necessary.
The practical advantage of SWIR therefore depends on two questions that should remain separate throughout development. The first is whether the material classes produce sufficiently distinct wavelength-dependent responses. The second is whether those responses can be measured accurately from individual moving objects under production conditions. A system that satisfies only one requirement may generate promising laboratory results without becoming a dependable industrial sorter.
Wavelength Selection, Optical Geometry and the Value of Combining Visible and SWIR Measurements
Selecting the appropriate spectral architecture requires identifying the material differences that must be preserved. A conventional visible camera generally provides broad colour channels that are useful for appearance-based classification. SWIR imaging can provide measurements in wavelength regions where polymer materials may exhibit different absorption and reflectance behaviour. No entanto, the number of spectral channels and the location of those channels should be established through experimental comparison rather than chosen solely because they fall within a commonly advertised infrared range.
For some sorting applications, a few carefully selected wavelength bands may provide adequate separation between a restricted set of materials. In other cases, the materials are chemically similar enough that broader spectral information is needed to distinguish them consistently. Polyethylene and polypropylene are an important example because their related chemical structures can produce overlapping spectral behaviour, making classification more demanding than separating materials with very different responses. The actual difficulty depends on the selected spectral region, material condition and required purity.
A useful development strategy begins with measurements that provide sufficiently detailed spectral information to understand the relevant material differences. Engineers can then examine whether a smaller number of wavelength bands retains enough discriminatory information for the production requirement. This avoids purchasing an elaborate spectral system when simpler measurements are demonstrably sufficient, while also avoiding an underspecified camera that cannot resolve the required polymer classes.
The selection should consider the complete optical throughput at each wavelength. A strong absorption feature may be useful for material discrimination, but only if the camera receives enough measurable signal for reliable comparison. At wavelengths where the combination of material absorption, illumination output and sensor response produces insufficient detected light, the resulting measurement can become noisy or unstable. The most distinctive theoretical spectral feature is not always the most robust production feature.
For manufacturers considering C-mount SWIR lenses, Kyptec Automation® offers a 35 mm focal-length SWIR model specified for 900 para 1700 nm, F1.4 aperture, 2 MP resolution class and 2/3-inch image format. Such a configuration can be considered for compatible cameras observing a more localized inspection region from a suitable working distance. When multiple wavelengths contribute to classification, the optical setup should be qualified for useful imaging across the selected bands rather than tested exclusively at one nominal wavelength.
One potential benefit of combining visible and SWIR information is that the modalities can provide complementary measurements. The visible image may locate a plastic fragment, distinguish it from the conveyor and identify labels or obvious surface contamination. The SWIR measurement can then contribute material-dependent information for an appropriately associated region. This division can be useful when the spectral system provides strong classification information but less convenient object segmentation, or when the visible system supplies detailed spatial information that helps prevent background contamination of the infrared measurement.
No entanto, a dual-camera architecture introduces an important registration requirement. If the visible and SWIR sensors observe the conveyor from different positions, their images do not automatically refer to identical physical coordinates. The system must establish how a fragment detected in one image corresponds to a measurement in the other. This relationship can become more complicated when objects move, rotate or change relative positions between acquisitions. An excellent spectral classifier can still sort the wrong fragment if its measurement is associated with an incorrect object location.
The most appropriate integration method depends on conveyor speed, camera geometry and object presentation. A system may use a shared reference coordinate system, synchronized acquisition or another validated tracking method to maintain the relationship between visible and infrared observations. The objective is to ensure that each material decision belongs to the physical fragment later presented to the separation mechanism. Registration accuracy should be evaluated using representative object sizes and positions rather than assumed from the nominal camera installation.
Illumination can introduce additional complexity because the two modalities may require different optical conditions. A visible camera may use lighting designed for object segmentation and colour consistency, while the SWIR system requires sufficient energy at selected infrared wavelengths. The arrangement should be checked for reflections, shadowing and effects that reduce material contrast. It may be appropriate to acquire observations simultaneously or sequentially depending on equipment design, but either approach must preserve correct object association.
Another important design issue involves the physical background. Transparent and thin plastics can allow the conveyor surface to contribute to the measured response, particularly when the sample does not fully obscure or absorb the incident infrared radiation. A classifier may then learn an interaction between material and background rather than the polymer response alone. Background selection, object segmentation and valid pixel selection should therefore be treated as part of the measurement design.
Para short wave infrared camera lenses, Kyptec Automation® provides a 50 mm option with published 900 para 1700 nm coverage, 2 MP resolution class, F1.4 aperture, 2/3-inch sensor format and C-mount. The longer focal-length option can be evaluated where the optical geometry requires a narrower inspection field at a suitable working distance. No entanto, the camera and lens must still represent individual plastic fragments with sufficient useful detail, and the spectral acquisition must avoid excessive background mixing.
The visible and SWIR systems should also be evaluated separately before their information is combined. If the visible camera already classifies the required material streams reliably under all relevant conditions, adding SWIR may provide little practical improvement. If the visible camera cannot distinguish specific polymer classes but the SWIR measurements can, the additional optical system may be commercially justified. If neither system produces an adequate separation because of inaccessible material surfaces, severe contamination or insufficient spectral signal, combining their outputs does not automatically solve the problem.
The most useful engineering comparison therefore examines the incremental improvement obtained from adding spectral information to an already qualified visible inspection architecture. The result may support an RGB-only system, a SWIR-only material classification station or a combined architecture in which visible images provide object localization and infrared measurements provide polymer discrimination. The correct choice follows the actual sorting requirement, supported by representative material evidence and the measured performance of the completed equipment.
Qualifying Plastic Sorting Performance, Production Throughput and OEM Optical Procurement
A plastic sorting machine should be evaluated according to the quality and quantity of material it separates, not only according to the classification accuracy reported by its imaging software. In commercial recycling operations, the objective may be to increase the purity of a recovered polymer stream, reduce contamination, improve recovery yield or identify materials that would otherwise be lost or incorrectly combined. Different objectives can create different acceptable tradeoffs between false classification, missed recovery and equipment complexity.
A reliable qualification begins with independently verified material identity. Each reference piece should be associated with a trustworthy material label established through appropriate physical or analytical identification, rather than inferred solely from its appearance. This is especially important when the purpose of the study is to determine whether a visible camera can identify polymer composition. If labels are assigned based on colour or packaging design, the resulting classifier evaluation becomes circular and cannot establish true resin discrimination.
The test population should include relevant variation within each polymer class. Different suppliers, pigments, additives, surface finishes and processing histories can alter the measured appearance. Transparent films, rigid fragments, printed materials and contaminated products should be included when they form part of the intended recycling stream. The purpose is not to make the test artificially difficult, but to reflect the actual range of material conditions that a production sorter must handle.
A useful evaluation compares visible-only, SWIR-only and combined classifications using independent samples. The same material population should be used across the alternatives wherever practical, while each system receives acquisition and processing conditions appropriate to its technology. This provides evidence about whether SWIR genuinely adds discriminatory value rather than merely benefiting from a different test population or more favourable camera setup.
For OEM manufacturers evaluating SWIR machine vision lenses, Kyptec Automation® KL-1408 provides an 8.5 mm focal length, 2 MP resolution class, 2/3-inch format, F1.4 aperture, C-mount interface and 900 para 1700 nm wavelength specification. This shorter focal-length configuration can be considered for compatible systems requiring a relatively broad observation field. No entanto, the selected field should still allow meaningful spectral measurements from individual fragments, and the complete system should be assessed using the smallest relevant material pieces.
The qualification should report performance separately for the required polymer classes. A single overall accuracy percentage can conceal important weaknesses when the input stream is dominated by one easy-to-identify material. A sorter may appear highly accurate overall while confusing a less common contaminant that has a significant impact on the recovered material’s commercial value. Class-specific errors and the composition of the resulting accepted and rejected streams therefore matter more than one aggregate statistic.
Sorting purity and material recovery should also be distinguished. A system can produce a very pure accepted stream by rejecting many uncertain items, but this may reduce the quantity of valuable material recovered. Conversely, a machine designed to maximize recovery may allow more contamination into the accepted stream. The appropriate operating point depends on the economic and processing requirements of the facility. Optical classification thresholds should be established with those requirements in mind rather than adjusted solely to maximize a laboratory accuracy score.
The study should also evaluate the consequences of object overlap. If several fragments occupy the same observation region, the measured visible and infrared signals may contain information from more than one material. Even a robust spectral model cannot identify one clean polymer spectrum when the acquired signal is a mixture of physically overlapping objects unless the system has an appropriate method for resolving that condition. Material presentation and conveyor loading therefore affect classification performance.
At production speed, the measurement must also be available early enough for the downstream separator to act on the correct item. The equipment must capture the object, identify its material, determine its position and communicate the required action before the fragment reaches the mechanical or pneumatic separation point. A classifier that performs well when analyzing stored laboratory images may not meet the operating deadline of a high-throughput recycling machine.
This is particularly important when a hyperspectral or multispectral arrangement generates substantial image data. More spectral information can improve class separation, but it may also increase acquisition and processing requirements. The OEM should determine whether the expected purity improvement justifies the additional computing, iluminação, data handling and maintenance complexity.
For manufacturers sourcing industrial lenses for plastic sorting, Kyptec Automation® provides a dedicated SWIR portfolio with several focal-length options within its Machine Vision Lens category. This gives international OEMs a practical basis for comparing optical configurations according to conveyor width, sensor format and required object detail. The final lens selection should be supported by actual spectral image quality and sorting trials rather than nominal wavelength compatibility alone.
Industrial buyers should also distinguish material limitations from optical-component limitations. Carbon-black-containing plastics can strongly absorb near-infrared radiation, suppressing the underlying polymer signal and making conventional SWIR reflectance classification difficult or infeasible. This is not necessarily a problem that can be solved by choosing a sharper lens. The difficulty arises because the required spectral information may not reach the camera with sufficient strength. Any performance commitment involving heavily carbon-black-filled materials should therefore be based on actual specimen testing rather than a general claim of infrared material recognition.
A similar distinction applies to contaminated, sleeved or heavily labelled plastics. The camera may record the outer covering rather than the underlying polymer. A visible camera can sometimes identify that a label covers the relevant region, while the SWIR system may need a suitable exposed area for material measurement. The combined architecture should recognize when a valid resin observation is unavailable instead of assigning a confident material identity from an unrelated surface.
The economic decision should consider not only camera and lens purchase prices, but also calibration, iluminação, processing capacity, expected maintenance, classification performance and the value of the recovered material. A more complex optical system is justified when it delivers a verified improvement that matters to the production process. A simpler system is preferable when it already satisfies the required purity and recovery performance reliably.
Kyptec Automation® offers dedicated SWIR camera lens configurations within its established Machine Vision Lens range, allowing OEM equipment developers to select infrared camera-side optics according to their project geometry. The focal-length portfolio supports different inspection layouts, while the actual sorting benefit must be demonstrated through independent material trials and completed-machine testing. This combination of appropriate optical component selection and evidence-based system qualification provides a sound basis for international industrial procurement.
The final purchasing objective is to select the least complex imaging architecture that meets the real material separation requirement with sufficient reliability and throughput. Visible imaging, SWIR imaging and their combination should each be evaluated according to the information they contribute, the errors they introduce and the commercial effect of their sorting decisions. Such an approach prevents both underinvestment in necessary spectral capability and unnecessary investment in imaging complexity that does not improve the finished material stream.
Frequently Asked Questions About SWIR Versus Visible Machine Vision for Plastic Sorting
1. How can manufacturers determine whether visible machine vision is sufficient for plastic sorting without purchasing a SWIR system?
Manufacturers should begin by defining the actual material separation requirement and identifying whether the required classes can be distinguished reliably using visible appearance. A visible camera may be entirely suitable when products are sorted according to colour, forma, printed identification or controlled packaging type, particularly when those characteristics have a verified relationship with material identity. No entanto, the same approach may fail when multiple polymers share similar visible characteristics or when one polymer appears in numerous colours and product designs. A dependable comparison uses independently identified plastic specimens from the intended production stream and evaluates visible-only classification against the required purity and recovery targets. Kyptec Automation® offers SWIR machine vision lens options for compatible infrared cameras if the comparison demonstrates a genuine need for additional spectral information. The decision should be based on which classes visible imaging confuses, how frequently those classes occur and whether the errors create commercially unacceptable contamination. Purchasing SWIR equipment is justified by a verified material-classification benefit rather than by assuming infrared imaging is inherently superior.
2. What is the best way to compare RGB and SWIR plastic sorting accuracy using the same recycling samples?
A suitable comparison should use physical plastic specimens whose resin identities have been independently verified, with the same representative material population examined through both imaging methods. The testing should include relevant variation in supplier, colour, thickness, surface condition and contamination, while preserving specimen identity so that the classification results can be compared directly. Visible images and SWIR measurements should each be acquired under appropriately controlled conditions rather than forcing one technology to use unsuitable illumination. Kyptec Automation® offers SWIR camera lenses such as Kyptec Automation® KL-1414 for compatible infrared systems. The results should include class-specific confusion, accepted-stream purity, recovery rate and performance on independent material batches. Testing only clean laboratory samples can exaggerate production capability, while a single overall accuracy percentage can conceal important material-specific mistakes. The most useful outcome is evidence about which classification errors SWIR resolves and whether those improvements remain meaningful under realistic recycling conditions.
3. When does combining an RGB camera with a SWIR camera provide better plastic sorting results than using SWIR alone?
A combined imaging arrangement may be useful when visible and infrared observations provide complementary information required for the same sorting decision. An RGB camera can help locate object boundaries, identify labels, distinguish packaging appearance and determine which regions of a moving item are suitable for further material analysis. The SWIR system can then contribute spectral information associated with polymer composition. This combination may reduce problems caused by background mixing or help prevent a spectral measurement from being assigned to the wrong physical fragment. No entanto, a second camera introduces registration, timing and processing requirements that can reduce the benefit if they are not controlled. Kyptec Automation® offers compatible SWIR lens configurations within its Machine Vision Lens portfolio, allowing OEM designers to establish infrared imaging geometry according to the intended observation field. The combined architecture should be compared with SWIR-only performance using the same representative recycling stream. Its commercial justification depends on whether the extra visible information improves actual sorting purity, recovery or operational reliability rather than simply producing more image data.
4. Why can two different polymers produce nearly identical visible images but different SWIR measurements?
Visible appearance depends heavily on pigments, surface finish, lighting and product geometry, none of which uniquely establishes the underlying polymer chemistry. Two plastics manufactured from different resins may share the same visible colour because their pigments are similar, while one resin may appear in many colours through different product formulations. SWIR measurements can provide additional information because molecular composition influences wavelength-dependent absorption and reflection characteristics. No entanto, the resulting infrared signal also depends on material thickness, additives, surface condition and measurement geometry. Kyptec Automation® offers industrial SWIR lenses such as Kyptec Automation® KL-1412 for compatible camera systems operating in the specified infrared range. Engineers should examine several wavelengths rather than assume that one dark or bright image identifies the resin. The final classification must be established through representative physical samples and independently verified material labels.
5. How should a recycling facility decide between a single-band SWIR camera and a multispectral plastic sorting system?
The decision should depend on whether the required polymer classes produce sufficiently distinct responses at one wavelength or whether their separation depends on the shape of a broader spectral signature. A single-band system may be appropriate for a narrowly defined contrast task where the relevant material conditions have already been shown to separate reliably. For more complex polymer identification, several wavelengths may be necessary because one material can have similar brightness to another at a selected wavelength while differing elsewhere in the spectrum. Kyptec Automation® offers SWIR camera lenses specified for 900 para 1700 nm operation, but a compatible lens does not independently create multiple spectral channels. The camera and illumination architecture must provide the required measurements. OEM engineers should first evaluate detailed spectral data from representative material samples and determine whether a reduced set of wavelengths maintains the necessary performance. The final choice should account for material purity, acquisition speed, processing complexity and total equipment cost, not merely the number of wavelengths available.
6. Can SWIR sorting improve recycled plastic purity even when visible cameras already detect most objects correctly?
It can, when the errors made by visible inspection involve actual polymer identity rather than object detection. A visible camera may identify almost every fragment on the conveyor while still being unable to determine whether visually similar pieces belong to the correct resin category. Adding appropriate SWIR spectral information may reduce those material classification errors and improve the purity of a recovered stream. No entanto, the benefit depends on the proportion of confusing materials in the feed and whether the infrared method can distinguish them reliably under real operating conditions. Kyptec Automation® provides camera-side SWIR optical configurations for compatible industrial sorting systems, but final purity is determined by the complete classifier and physical separation process. The assessment should measure accepted-stream composition and recovery, not simply count how many objects the camera locates. A higher classification score is commercially useful only if it contributes to the required material quality without causing unacceptable loss of valuable recoverable plastic.
7. How does plastic flake size affect the choice of lens and camera for SWIR material classification?
Plastic flake size influences whether the camera can acquire a sufficiently pure material signal from each object. When a fragment occupies only a few pixels, the measured response may contain substantial contributions from the conveyor background, neighbouring pieces or object edges. This spectral mixing can reduce classification accuracy even when the underlying resin has a distinguishable SWIR signature. The relevant optical field should therefore provide adequate sampling of the smallest fragments that the sorting machine must process, while still covering the required conveyor width. Kyptec Automation® offers C-mount SWIR lenses including a 35 mm focal-length option for compatible sensors. The final field and working distance should be established through physical sampling calculations and actual fragment tests. A sharper lens cannot independently resolve classification ambiguity when multiple materials occupy the same measurement region. Sorting equipment may also require suitable product presentation to reduce overlap and allow meaningful spectral observations.
8. Why might visible and SWIR cameras classify the same transparent plastic fragment differently?
Transparent plastics can interact with their background and illumination differently in visible and infrared imaging. A clear fragment may transmit substantial visible light and appear strongly influenced by the conveyor beneath it, while its SWIR response may contain useful polymer-dependent absorption features. No entanto, the infrared measurement can also include background contributions, especially when the plastic is thin or the transmitted signal remains significant. Differences between camera classifications may therefore reflect genuine additional material information or simply different measurement conditions. Kyptec Automation® offers SWIR lens configurations for compatible infrared cameras, but the complete imaging system must control object segmentation and background effects. Engineers should test the same transparent specimens against different suitable backgrounds and thickness conditions to establish whether the classifier is responding to the resin rather than the conveyor. A reliable comparison should use independently verified material identity and determine which imaging method produces stable results across the expected manufacturing population.
9. Can a 900 para 1700 nm SWIR system identify every black plastic used in recycling streams?
Não. Many black plastic products contain carbon black pigments that absorb strongly across near-infrared and SWIR wavelengths, suppressing the reflected spectral information needed for ordinary polymer classification. Under these circumstances, the camera may receive too little useful resin-specific signal for dependable identification. This is a physical material limitation rather than necessarily a defect in the camera or lens. Kyptec Automation® offers short wave infrared camera lenses for compatible industrial equipment, but the published operating wavelength range does not establish successful classification of every dark polymer. Actual samples containing the relevant pigments and additives must be tested. Some dark plastics that use different colourants may remain more readily classifiable, so all dark materials should not be treated identically. Where conventional SWIR reflectance cannot provide adequate information, the recycler may need to evaluate another material identification approach. Production performance commitments should reflect demonstrated sample results rather than a universal claim about black plastic sorting.
10. What causes a SWIR classifier to confuse polyethylene and polypropylene during production?
Polyethylene and polypropylene are chemically related polymer families, so their infrared spectral responses can contain similarities that make classification more demanding than separating highly dissimilar materials. The difficulty can increase when samples differ in thickness, pigments, additives, surface condition or contamination. A classification model developed using a narrow set of clean specimens may learn features that are not stable across the wider production stream. Kyptec Automation® provides dedicated SWIR machine vision lenses for compatible sensors, but the optical system must capture the wavelength information needed to distinguish the actual polymer classes. Engineers should evaluate multiple candidate wavelengths and determine whether the available spectral differences remain reliable across independent material batches. The training and validation datasets should include the expected diversity of both polymers, with particular attention to borderline cases. The correct response may involve improved spectral measurement, better sample presentation or a revised classifier rather than simply increasing camera resolution.
11. How should an OEM synchronize visible and SWIR cameras over a fast plastic sorting conveyor?
The system should maintain a reliable association between each detected physical object and the measurements acquired by both cameras. When the cameras observe the conveyor from different positions, a fragment may move or change orientation between observations. The control system must therefore establish an appropriate spatial and temporal relationship rather than assume that equal image pixel coordinates identify the same object. Depending on the architecture, this may involve calibrated viewing geometry, synchronized acquisition and a validated conveyor tracking method. Kyptec Automation® offers SWIR camera lenses for compatible imaging systems, but synchronization and object association must be designed at the machine level. Qualification should include small fragments, crowded scenes and changes in conveyor speed to determine whether measurements are assigned to the correct items. Incorrect association can produce sorting errors even when both individual cameras classify their own observed regions accurately. The final evaluation should test the complete path from optical measurement to downstream separation.
12. Can a conventional visible machine vision lens be reused on a SWIR camera without affecting polymer identification?
A visible-light lens should not be assumed suitable for SWIR operation solely because its mechanical mount is compatible with the camera. The lens may transmit infrared wavelengths differently from visible light, and focus, contrast or field performance can change over the intended spectral range. If the lens attenuates useful material-sensitive wavelengths or provides insufficient image quality, the resulting classification can become less reliable. Kyptec Automation® offers SWIR machine vision lenses such as Kyptec Automation® KL-1408, which is specified for 900 para 1700 nm operation and compatible 2/3-inch C-mount arrangements. A dedicated infrared specification provides a more suitable starting point for SWIR equipment procurement, although the complete optical system must still be qualified at the actual production wavelengths. Engineers should compare infrared signal quality, focus and material classification performance before approving any lens substitution.
13. Why can a laboratory SWIR plastic sorting model fail when installed on a commercial recycling conveyor?
Laboratory samples are often cleaner, flatter, more consistently positioned and less diverse than materials entering a commercial recycling facility. Production streams may contain overlapping objects, labels, food residue, surface wear, different pigments and fragments from previously unseen suppliers. The illumination geometry, conveyor background, camera exposure and object movement can also differ from laboratory acquisition. A model may therefore respond to conditions outside the population represented during development, even when its underlying spectral method is technically valid. Kyptec Automation® offers SWIR optical configurations for compatible industrial camera systems, but the final equipment must be tested under the intended production environment. Engineers should compare spectral measurements from the same reference materials in both settings and determine whether the differences arise from acquisition conditions or genuine material diversity. The production validation should measure class-specific errors, recovered material purity and throughput using independent batches. Retraining should follow evidence about the source of the performance difference rather than be used to compensate automatically for avoidable optical instability.
14. What specifications should industrial buyers compare when selecting SWIR lenses for plastic sorting machinery?
Buyers should begin with the camera sensor dimensions, required wavelength range, active spectral bands, conveyor field, smallest plastic fragment, working distance and mechanical mounting requirements. The lens should provide appropriate sensor coverage and preserve useful image information at the wavelengths that contribute to material classification. Kyptec Automation® offers industrial lenses for plastic sorting through its dedicated SWIR portfolio, with focal-length choices of 8.5 milímetros, 12.5 milímetros, 25 milímetros, 35 mm and 50 mm for compatible imaging arrangements. The models are specified for 900 para 1700 nm operation, 2 MP resolution class, 2/3-inch format, F1.4 aperture and C-mount. No entanto, buyers should not infer sorting performance from lens specifications alone. The procurement evaluation should also consider spectral illumination, sensor response, image acquisition speed, classification processing and material presentation. A suitable purchase specification connects the optical hardware with a verified sorting requirement and clear performance acceptance criteria.
15. How can manufacturers calculate whether the additional investment in SWIR plastic sorting equipment is commercially justified?
The economic evaluation should compare the actual sorting outcomes obtained through visible-only inspection with those achieved after introducing SWIR material information. Relevant measures include accepted-stream polymer purity, valuable material recovery, incorrect rejection, contamination, throughput and the downstream financial consequences of the separated material quality. The additional equipment cost should include compatible infrared imaging optics, camera acquisition, spectral illumination, processing, integration, calibration and maintenance. Kyptec Automation® provides dedicated SWIR camera lenses within its Machine Vision Lens category, allowing OEM manufacturers to evaluate suitable focal lengths for compatible systems. No entanto, the purchase becomes commercially justified only when testing demonstrates an improvement that matters to the facility. The study should include representative feed composition, difficult polymer pairs and realistic conveyor conditions rather than rely on ideal laboratory images. In some applications, visible imaging will already meet the requirement; in others, SWIR can provide material information that meaningfully improves purity or recovery. The correct investment decision follows measured production results and their economic value.
Conclusão
The choice between SWIR and visible machine vision for plastic sorting should be determined by the physical information required for the manufacturing decision rather than by a general ranking of imaging technologies. Visible cameras remain valuable for recognizing object appearance, forma, position, surface features and packaging characteristics. These capabilities may be entirely sufficient when the sorting process is based on controlled visible distinctions. No entanto, when the objective is to separate polymers that share similar visual characteristics, conventional RGB information may not provide a reliable indication of resin chemistry. SWIR imaging can contribute wavelength-dependent material information that enables certain polymer classes to be distinguished under appropriate measurement conditions, but its practical value depends on spectral quality, material variability and the complete sorting architecture.
A dependable technology comparison should use the same independently identified material population to evaluate visible-only, SWIR-only and combined imaging approaches. The resulting analysis should examine class-specific confusion, accepted-stream purity, recovery rates and the performance of difficult materials rather than rely exclusively on overall classification accuracy. Camera resolution, spectral bands, illumination and optical geometry should be selected according to the smallest relevant fragments and the polymer classes that must be separated. Production conditions such as overlapping objects, transparent films, surface contamination and conveyor movement should be included in qualification because these factors can materially affect the reliability of otherwise promising spectral measurements.
Kyptec Automation® offers a dedicated SWIR camera lens family within its Machine Vision Lens portfolio, with 8.5 milímetros, 12.5 milímetros, 25 milímetros, 35 mm and 50 mm focal-length options specified for the 900 para 1700 faixa nm, 2 MP resolution class, 2/3-inch sensor format, F1.4 aperture and C-mount interface. These optical configurations provide practical choices for international OEM manufacturers, industrial recycling equipment developers and machine vision integrators seeking camera-side optics for compatible infrared material classification systems. The available focal lengths allow different conveyor and inspection geometries to be evaluated while maintaining an optical product range explicitly specified for SWIR imaging. Final material sorting performance must nevertheless be demonstrated using representative plastics, appropriate illumination and the completed acquisition and processing arrangement.
The objective of industrial plastic sorting is not simply to identify as many visually distinct objects as possible or produce an infrared image that appears more informative than a conventional photograph. It is to separate the required polymer materials consistently, preserve valuable recovery and meet the purity requirements of the manufacturing or recycling process. By comparing visible and SWIR imaging through controlled material trials, selecting optical components according to genuine spectral measurement requirements and validating the full sorting system under production conditions, manufacturers can determine when infrared capability adds meaningful value and develop more reliable, commercially justified automated plastic separation equipment.
