Clear
Blue
Green
White
Red
High accuracy
Consistent results with minimal false calls
Technologies
Sorting technologies explained as part of one production system.
Every Mayson sorter is configured from the inspection layers its material actually needs — visible-light imaging, AI-assisted recognition, multi-view inspection, lighting-and-ejector control, and near-infrared (NIR) material recognition. Not every machine uses all of them: the right combination is selected for the stream, the reject classes, and the line it runs in. Each section below sets out what the technology can see, where it fits, and how the result is confirmed on your own material.
How they work together
Lighting, imaging, recognition, and ejection are planned as one sorting path.
Each technology page explains what the inspection layer does, where it fits, and how it connects to the products and materials it supports.
- 01FeedMaterial is metered into an even, stable flow.
- 02PresentChute or belt presents particles to the view.
- 03IlluminateTuned lighting makes target defects visible.
- 04CaptureCameras image every particle in position.
- 05RecogniseRecipe and AI logic classify accept vs reject.
- 06EjectTimed air pulses remove the rejected stream.
Optical technology
Visible-Light Inspection
Visible-light analysis classifies particles by color, transparency, contour, gloss, surface texture, and visible defects.It does not confirm polymer identity by itself — material decisions use NIR or multi-spectral recognition. High-speed inspection supports recipe-based classification and precisely timed rejection.
Live material stream
RunningThroughput 3,842 pcs/min
ID 04217Class: RedConf: 99.2%
Analysis dashboard
Color segmentation
Clear
Blue
Green
White
Red
RGB distribution
Channel values
R212
G46
B48
Contour and shape
- Area
- 24.7 mm²
- Perimeter
- 21.3 mm
- Circularity
- 0.68
- Aspect ratio
- 1.23
ClassificationRed
Confidence99.2%
✓ Accept for red streamSpectral signature
Color space (CIE LAB)
- L*
- 52.6
- a*
- 68.3
- b*
- 43.1
Maximum uptime
Stable performance in demanding environments
Adaptive recipes
Programs tuned to real material samples
Traceable and repeatable
Data-rich insight for quality control
Intelligent recognition
AI-Assisted Recognition
Recipe-linked models work alongside conventional sorting rules to recognise subtle, variable, or hard-to-parameterise defect classes with greater accuracy.
How it works
- 01Define from real data
your team builds defect categories from real sample images and production experience, not generic presets.
- 02Configure recognition logic
recognition rules and model-assisted logic are tuned around the target material and its reject classes, so each decision stays explainable.
- 03Validate before scaling
recognition accuracy is confirmed on representative samples before it drives production decisions, and the model assists the operator rather than replacing human judgement.
What it identifies
Complex visual defectsMixed appearance classesProduct-specific reject groups
Suitable materials
Application scope
- Training examples and sample variation must be reviewed before any application-specific claim is published.
- AI wording must not imply autonomous quality decisions made without operator validation.
- A material test should confirm whether the defect categories are stable enough for production.
Inspection architecture
Multi-View Inspection
Multiple synchronised viewpoints reveal surfaces, edges, and orientation-dependent defects that a single camera angle can miss — building a more complete understanding of every product.
How it works
- 01Configure viewpoints
cameras are positioned around the product so each angle exposes a different visible surface, matched to its geometry and flow.
- 02Correlate observations
views are combined into a surface-visibility map so defect cues are cross-checked across angles, not judged from one.
- 03Confirm critical surfaces
a part is only classified once the surfaces that matter are visible and the confidence threshold is met.
What it identifies
Side-specific visible defectsFine cracks, chips, and seam defects on edgesSurface marks exposed only by orientation
Suitable materials
Application scope
- It is a configuration pattern, not a universal claim for every product family.
- Stable material flow and consistent particle spacing and rotation must be tested.
- Performance metrics require verified test records for the specific product.
Control layer
Lighting and Ejector Control
The practical control layer that synchronises inspection visibility, decision timing, and reject actuation for consistent, accurate sorting.
How it works
- 01See clearly
precision lighting reveals target defects and reduces visual ambiguity across different materials.
- 02Time precisely
detection, particle position, and actuation are synchronised to the millisecond so the right pulse meets the right particle.
- 03Eject accurately
compressed-air actuation is tuned to lift the rejected stream out cleanly without disturbing accepted product.
What it identifies
Recipe timing windowsReject actuation timingLighting-sensitive visible defects
Suitable materials
Application scope
- Hardware details must match the actual product configuration.
- Compressed air, product speed, and material spread all affect final rejection behaviour.
- Recipes must be documented per material test, not copied across unrelated streams, and reject-accuracy claims need approved test data.
Spectral technology
NIR Inspection Technology
Near-infrared (NIR) analysis identifies polymer type by measuring spectral response across multiple wavelengths — enabling high-speed, accurate plastic sorting.
How it works
- 01Polymer signature recognition
each polymer has a distinct spectral fingerprint that is captured across the near-infrared range.
- 02Multi-wavelength analysis
broad spectral scanning separates plastics that look identical to the human eye or a colour camera.
- 03Real-time classification
within milliseconds each particle is classified and routed, so sorting keeps pace with production speed.
What it identifies
Polymer type by spectral signature (PET, HDPE, PP, PE, PVC, PS, ABS, PC)Additives, coatings, and fillers that shift the signalComposition conflicts such as moisture, ageing, or contamination
Suitable materials
Application scope
- NIR cannot reliably identify black or carbon-black plastics, which absorb the signal.
- Best performance is on light-coloured and translucent plastics with clean, dry surfaces.
- Accuracy depends on validated spectral libraries, stable flow, and periodic calibration.
Technology comparison
Visible-light vs NIR vs AI: what is the difference?
Each layer answers a different question about the same particle. Most Mayson sorters combine two or more layers, and the right mix is confirmed with a sample-led material test.
| Technology | Sees | Does not do alone | Typical use |
|---|---|---|---|
| Visible-light inspection | Color, transparency, contour, gloss, surface defects — including black pieces | Does not confirm polymer type | MAS-C, glass sorter, food sorters, MAS-B appearance classes |
| NIR material recognition | Polymer spectral response (PET vs PVC, PE, PP…) | Cannot reliably identify black plastics; does not replace color sorting | MAS-P, MAS-PC |
| Multi-spectral polymer recognition | Harder polymer spectral differences (PP vs PE, ABS, PC, PMMA…) | Still cannot reliably identify black materials | MAS-P Pro (MAS 3D), MAS-PC Pro |
| AI-assisted recognition | Complex appearance classes — bottle shape, labels, brands, deformation, anomalies | Is not an autonomous quality guarantee | MAS-B, complex food and agricultural appearance targets |
| Multi-view inspection | Multi-angle surface, edge and orientation-dependent defects | Is not required for every material | Nuts, beans, seeds, some complex-appearance materials |
| Lighting and ejector control | Visibility, timing and air-jet rejection | Is not a recognition technology by itself | Every chute and optical sorting system |
Buyer questions
Sorting technology FAQ
What is visible-light inspection?
Visible-light inspection uses high-resolution cameras and tuned lighting to classify particles by color, transparency, contour, gloss, surface texture and visible defects. It powers MAS-C, the glass sorter, food sorters and MAS-B appearance classes. It does not confirm polymer identity by itself.
What is NIR material recognition?
NIR (near-infrared) recognition reads each particle's polymer spectral response to identify material type — for example separating PET from PVC, PE or PP regardless of color. It powers MAS-P and MAS-PC. Its main limit: black and carbon-black plastics absorb the signal and cannot be reliably identified.
What is AI-assisted recognition?
AI-assisted recognition is deep-learning appearance classification trained on real samples. It handles complex categories that fixed rules cannot — bottle shapes, labels, brands, deformation and anomalies — and powers MAS-B and complex food and agricultural targets. It supports operator-validated decisions; it is not an autonomous quality guarantee.
Can a color sorter identify polymer type?
No. Visible-light color sorting classifies what the camera can see — color, transparency and visible appearance. A flake with the right color can still be the wrong material. Confirming polymer identity requires NIR recognition (MAS-P) or multi-spectral recognition (MAS-P Pro), which read material response instead of color.
Can NIR sort black plastic?
NIR cannot reliably identify black or carbon-black plastics because they absorb the near-infrared signal. Black pieces can still be removed as a visible-color reject class by a color sorter such as MAS-C — but removing black is not the same as identifying black material by polymer type.
Does AI replace material testing?
No. Every Mayson configuration is confirmed with a sample-led material test before performance expectations are set. AI recognition is trained and validated per application on your real material, so the test is what proves whether the defect categories are stable enough for production.
Material test
Technology fit is defined by the material, not the brochure.
Send a representative sample and we will show which inspection path best fits the material, output target, and production line.