ISO 9001 Certified Manufacturer 360° Factory Tour

Rice Sorting | Visible and Infrared Spectral Selection

Rice Color Sorter Filters

Optical filtering for visible colour recognition and infrared foreign-material detection

Rice color sorter filters control the spectral range entering the imaging system. Working with the illumination source, industrial camera and image recognition algorithm, they reveal differences in grain colour, brightness and reflection or transmission at selected wavelengths. For visible colour sorting, near-infrared assistance and short-wave infrared material detection, IR-cut, bandpass, longpass and dichroic elements help stabilise the image signal used for discoloured grain recognition, appearance grading and selected foreign-material detection.

  • Visible colour sorting
  • NIR-assisted detection
  • SWIR detection
  • Line-scan imaging
  • Spectral and mechanical customisation
400-700 nm Visible colour channel evaluation range
850 / 940 nm NIR LED evaluation points
900-1700 nm Standard InGaAs line-scan response
Rice sorter inspection zone with optical filter elements
Source, grain, filter and camera positions are schematic and vary between machines

What is a rice color sorter filter?

A rice color sorter filter is an optical element installed at the illumination or imaging end of sorting equipment. By selectively transmitting, reflecting or absorbing different wavelengths, it adjusts the spectral composition that actually reaches the camera.

When illuminated, grains and contaminants produce different reflection, transmission and absorption signatures. Some differences are visible as colour changes such as yellow grains, black spots and bran specks; others may be more pronounced in the near-infrared or short-wave infrared. The filter retains the spectral signal relevant to the detection task while suppressing non-target bands that would disturb imaging.

Two types of requirement should be separated. Colour detection retains fairly complete visible information and suppresses unwanted infrared response so the colour camera can render grain colour consistently. Specific-band detection selects one or more bands and compares the image differences between grains, background and foreign material, giving the classification algorithm extra information.

Commercial sorters already combine RGB colour imaging with infrared detection, but sensor and optical layouts differ between machines.

How does a filter differ from an RGB camera?

An RGB camera obtains red, green and blue information through its own colour sampling structure; a filter mounted in front of the lens or inside the optical path further controls the overall incident spectrum. Colour detection systems normally do not need three separate red, green and blue filters in front of the lens.

An overly narrow single-band bandpass filter may instead discard the information needed to judge grain colour.

A filter only adjusts the spectrum entering the imaging system. It does not replace the camera colour sampling structure and does not by itself determine sorting results.
Illumination Imaging Light source Broadband Grains and contaminants Filter Selects spectrum Camera / line-scan sensor Image algorithm Classify / eject ※ Order is schematic; the filter may sit at the source or inside the lens Design schematic, not measured product data
A filter only adjusts the spectrum reaching the camera; sorting also depends on feeding, algorithm, calibration and ejection (schematic, not measured data)

Main Application Scenarios

Different tasks call for different observed features and optical schemes, and should be assessed against real samples.

Detection task Observed features Optical scheme to evaluate Selection focus
Yellow and discoloured grains Overall or local grain colour change Visible imaging with an IR-cut filter Preserve colour differences, control IR crosstalk
Black spots, bran specks and surface defects Local colour, brightness and texture changes Broadband visible imaging; a monochrome auxiliary channel if needed Resolution, illumination uniformity and spot contrast
Chalky grains and translucency differences Scattering, transmission and brightness differences Filter scheme matched to reflection or transmission layout Source direction, background and exposure conditions
Broken grains and shape grading Length, contour and integrity Filter matched to the contour-imaging source The filter stabilises imaging; size judgement is done by imaging and algorithm
Selected same-colour foreign material Spectral differences not obvious in visible light NIR or SWIR detection channel Compare real grain and contaminant spectra or images first
Multi-channel combined detection Colour, shape and infrared features together Visible/IR splitting, or separate filters on multiple cameras Channel isolation, image registration and signal strength
Common rice sorting tasks include yellow grains, grey grains, bran specks and broken grains; the applicable scheme depends on the raw material and equipment configuration. For glass, plastic and stone contaminants, adding an infrared filter alone does not guarantee detection. Material, colour, thickness, size and illumination all affect the result.
Visible channel: colour and surface differences NIR channel: some material differences may be clearer Normal Normal Yellow Yellow Black spot Black spot Bran speck Bran speck Contaminant Contaminant Compare real samples first The same contaminant may differ by band or closely resemble grain; validate with representative samples, then add channels or change illumination if needed. Design schematic, not measured product data; sample forms and grey levels are illustrative
Tasks must be compared with real grain and contaminant samples; band numbers alone do not determine detection (schematic, not measured data)

Filter Types and Band Selection

Choose the filter structure by detection channel, after confirming the sensor, light source and real incidence conditions.

Type A

Visible imaging filters

For channels where grain colour, surface spots and appearance differences drive recognition. The priority is to retain the target visible range while controlling ultraviolet and near-infrared influence on the image. Where visible and infrared illumination are both present, IR suppression in the colour channel must be assessed together with the camera existing filter structure to avoid duplication or a mismatched cut edge.

Type B

NIR bandpass filters

For a separate detection channel paired with a near-infrared source, retaining the useful spectrum around the emitter and reducing crosstalk from visible light or other illumination channels. 850 nm and 940 nm are evaluation points for matching LEDs; confirm that real samples show sufficient difference at those bands before selecting.

Type C

SWIR filters

For detection systems fitted with a suitable short-wave infrared sensor, using bandpass or longpass designs chosen from material spectral differences. Some standard InGaAs line-scan sensors respond over roughly 900-1700 nm; that is the sensor response range, not a band that any single filter must transmit in full.

Type D

Visible/IR dichroic elements

For routing different bands into different detection channels. Transmission band, reflection band, mounting angle and polarisation must all be specified, and the design suits equipment with a defined dual-channel layout.

850 nm and 940 nm are evaluation points for matching LEDs, not the optimum detection bands for every grain and contaminant combination. Confirm sufficient sample contrast at those bands first. A standard silicon camera cannot be turned into a short-wave infrared camera by changing the filter alone.
Channel and band evaluation ranges (schematic) 400 700 850 940 900 1700 Wavelength / nm (not to scale) Visible colour channel Approx. 400-700 nm 850 nm 940 nm Evaluation points for matching LEDs InGaAs sensor response (not a filter passband) A silicon camera cannot be upgraded to 900-1700 nm by changing the filter; SWIR needs a matched sensor, lens and illumination.
Bands are evaluation ranges, not a fixed link between one wavelength and one contaminant (schematic, not measured data)

Channel Schemes and Supporting Requirements

Bands are evaluation ranges and must be confirmed together with the sensor, lens and illumination.

Channel or scheme Band reference Main purpose Supporting requirements
Visible colour channel Approx. 400-700 nm, edges set by the system Acquire colour and appearance information Colour camera and visible illumination
NIR channel near 850 nm Passband covering the actual LED emission spectrum Acquire reflection or transmission images in that band Camera with matching response and an NIR lens
NIR channel near 940 nm Passband covering the actual LED emission spectrum Evaluate sample contrast in a second NIR channel Check camera response and light budget at short exposure
Selective SWIR channel Selected within the usable sensor response range Acquire material-related spectral differences Matched InGaAs sensor, source and lens
Visible/IR splitting Boundary set by the two channel detection ranges Acquire different bands independently Design and inspect at the real angle of incidence
These bands are evaluation ranges and do not imply a fixed link between one wavelength and one contaminant. SWIR detection needs a matched sensor, lens and illumination; a filter cannot change the response range of the sensor itself.

Published Product Parameter Reference

The following parameters come from published machine-vision filter brand literature. They illustrate how specifications are expressed, and do not represent verified rice sorting performance or a supply commitment from us.

Reference product Filter type Manufacturer stated usable range FWHM Peak transmittance Other parameters
SP700 Visible pass, UV/NIR suppression 405-690 nm Not applicable to the bandpass definition here >=90% 50% transmittance edges at 400/700 nm, stated tolerance +/-15 nm
BP850 NIR bandpass 820-910 nm 160 nm >=90% For 840/850 nm LEDs
BP940 NIR bandpass 928-955 nm 55 nm >=85% For 940 nm LEDs, stated tolerance +/-10 nm
Usable range, FWHM and peak transmittance describe different metrics and cannot be substituted for one another. For example, a usable range of 820-910 nm does not mean the FWHM is 90 nm. A peak transmittance of 90% does not mean the whole working band reaches 90%; custom acceptance should state the minimum or average transmittance within the passband.
Published specification schematic (not our measured curves) T% Wavelength / nm (not to scale) 400 700 850 940 1200 1700 90% SP700 405-690 nm BP850 820-910 nm BP940 928-955 nm Usable range is not FWHM; peak transmittance does not apply across the whole band.
Curves illustrate how specifications are expressed; not verified for rice sorting and not a supply commitment. Design schematic, not measured data

Key Technical Parameters and Definitions

Parameters must be matched to the source emission spectrum, sample differences and sensor response.

Parameter Definition Impact on rice sorting equipment
Centre wavelength CWL Position of a bandpass filter passband. For asymmetric spectra the calculation method must be agreed; do not judge from the product name alone Must match the source emission spectrum, sample differences and sensor response
Full width at half maximum FWHM Wavelength interval between the two half-peak points under an agreed baseline A narrower band usually rejects more unwanted light but can also reduce useful signal
Cut-off / cut-on wavelength Spectral edge of a shortpass or longpass filter, usually defined at a stated transmittance such as 50% Determines which wavelengths enter the channel, affecting colour information or IR channel isolation
Passband transmittance How much light in the target band passes the filter; peak, average and minimum values must be distinguished Affects signal strength at short exposure and the required source power
Blocking depth OD Suppression of unwanted light, OD = -log10(T) where T is the transmittance fraction; OD3 = 0.1%, OD4 = 0.01% Must be specified together with the blocking band; writing OD4 alone does not define performance
Angle of incidence AOI Angle between the ray and the filter surface normal Interference filter spectra shift with incidence; design against lens field of view and beam angles
Spectral spatial uniformity Consistency of passband, edge and transmittance across the filter Especially important for long strip filters and wide line-scan imaging to avoid position-dependent response
Surface quality Requirement for scratches and digs, with a stated evaluation standard Defects may add scatter or affect local image areas; actual impact depends on mounting position
Flatness and transmitted wavefront Flatness describes surface shape, transmitted wavefront describes wavefront change after passing through; the two are not equivalent Set the requirement according to where the filter sits in the imaging path
Clear aperture Area meeting optical requirements and not obscured by the mount Must cover the real imaging beam to avoid mechanical vignetting
Wavelength range and operating conditions Nominal wavelength, real AOI range, temperature and polarisation state Validate under real running conditions; a single room-temperature curve does not replace machine performance
Narrower is not automatically better. Grains move fast and exposure time is limited. An overly narrow passband can clip part of the LED spectrum and leave the image signal too weak. A sound design balances stray-light rejection against useful light throughput.
Bandpass curve definitions (design schematic, not measured data) T% Wavelength / nm Half peak 50% Peak / CWL FWHM Cut-on edge Blocking: OD needs a stated band OD3 = 0.1%, OD4 = 0.01%; writing OD4 alone does not define performance.
Parameter definitions and acceptance wording; measured curves must state model, AOI and test conditions (schematic, not measured data)

Custom Fabrication and Deliverables

This section lists items that can be evaluated per project. Size ranges, tolerances and performance limits should be published only after fabrication and inspection verification.

Custom direction Points to specify
Spectral customisation Centre wavelength, bandwidth, cut edge, minimum/average passband transmittance, blocking band and blocking depth
Optical path adaptation Mounting position, nominal AOI, AOI range, beam convergence and polarisation state
Substrate selection Optical glass, fused silica or other materials assessed against band, transmission, mechanics and environment
Outline and structure Round, rectangular or long strip form, dimensions, thickness, chamfers and edge requirements
Surface and coating Surface quality, coating uniformity within the clear aperture, cleaning method and durability
Mounting adaptation Bare, retaining ring or frame mounting, orientation, edge obscuration and mounting stress
Batch consistency Spectral and dimensional deviations across pieces and batches, and agreed inspection positions
Inspection deliverables Spectral curve, test angle, sample ID, dimension record and agreed inspection report
Machine validation should cover real running speed, grain attitude, raw material batches and typical contaminant samples. Contrast improvement in a single static image cannot be equated with sorting performance in production.
Common fabricated forms (schematic) Round (bare / ring) Rectangular (mounted) Strip (line-scan) Outline, thickness, chamfer, clear aperture and coating uniformity follow drawing and inspection; forms are schematic, not product photos. Strip filters in particular need specified spectral uniformity across the aperture
Size ranges, tolerances and performance limits are published after fabrication and inspection; schematic, not measured data

Figures and Content Notes

Diagrams explain where the filter sits and what the parameters mean; they do not replace measured data.

Page position Suggested material and information to convey
Hero Rice sorter inspection zone combined with real filter parts, showing the application context
Definition Source - grain - filter - camera schematic showing where the filter sits
Applications Normal, yellow, black-spot and bran-speck grains plus typical contaminants for different tasks
Band selection Visible, NIR and SWIR channel schematic separating channels from sensor requirements
Specifications Measured transmittance curves with annotated bandwidth, cut edge and transmittance
Custom capability Round, rectangular and long strip product photos showing real fabricated forms
Comparison images should use the same sample batch and state illumination, camera and exposure conditions; design curves should be labelled schematic, and measured curves should state the test conditions.

Recommended Custom Process

Requirement review > Sample and source matching > Spectral and mechanical design > Sample fabrication > Optical inspection > Machine trial > Batch specification sign-off

  1. 01

    Requirement review

    Define the detection task, target contaminants, optical layout and running speed.

  2. 02

    Sample and source matching

    Compare candidate bands using representative grain and contaminant samples.

  3. 03

    Spectral and mechanical design

    Set centre wavelength, bandwidth, cut edge, transmittance and outline.

  4. 04

    Sample fabrication

    Complete substrate, coating and edge work to drawing.

  5. 05

    Optical inspection

    Check spectrum, uniformity across the aperture, dimensions and appearance.

  6. 06

    Machine trial

    Verify imaging at real running speed, grain attitude and illumination.

  7. 07

    Batch specification sign-off

    Form the product drawing, acceptance conditions and batch consistency requirements.

Frequently Asked Questions

Source Factory In-house coating lines, no middlemen
Ready to Ship Standard items dispatched same or next day
Quality Guaranteed 100 % inspected, spectral report included
Express Delivery DHL / FedEx worldwide