Direct answer: Choose an optical sorter by the defect you must remove, not by the machine’s headline capacity. Match the sensor to the defect (visible colour, near-infrared or shape), set capacity at the purity and good-product loss you can accept, and confirm both with a trial on your own material. Optical sorting reduces risk but does not replace testing.
Key takeaways
- Write a defect list first: what must be rejected, how it looks, and how much of it arrives in the feed.
- Visible-light (colour) cameras see surface colour. Near-infrared (NIR) sensors can detect some differences that the eye cannot. Shape software helps with broken pieces and foreign objects.
- Throughput, purity and the amount of good product thrown away are linked. Raising one usually costs another.
- A second pass (re-sort) on the reject stream often recovers good product. Plan conveyors for it.
- Food-safety claims such as aflatoxin reduction must be verified by laboratory testing of the sorted stream.
What does an optical sorter actually do?
Short answer: An optical sorter spreads product into a thin stream, views each piece with cameras or sensors, and fires short air jets (ejectors) to blow out pieces that fail the set rules. It is a fast, automatic inspection and rejection step. It works on what it can see, so the defect must look different from good product.
Most machines follow the same sequence: a vibratory feeder, a chute or belt that presents product in a single layer, a viewing zone with lights and sensors, a processor that compares each piece with the set rules, and an ejector bank. Accepted product and rejects fall into separate outlets.
On GTI’s sorting machine page, optical sorting is listed with vibratory and roller-based sorting, in single or multi-stage layouts. Size grading is a different job. If you need to split product by size, see our guide to vibratory, rotary and roller graders.
Which sensor suits which defect?
Short answer: Use colour cameras for defects that look different to the eye, such as black, discoloured or mouldy pieces. Consider NIR where defects and good product look similar but differ in composition or internal structure. Use shape analysis for broken pieces, clusters and some foreign bodies. Many lines need more than one.
| Defect type | Examples | Sensor usually considered | Watch-outs |
|---|---|---|---|
| Colour difference visible to the eye | Black or burnt grains, discoloured pulses, dark chilli pods | Visible colour cameras (one or two sides) | Natural colour variation can cause false rejects |
| Similar colour, different material | Stones, glass, plastic or shell pieces in nuts and seeds | NIR or combined visible + NIR | Must be proven on your own foreign matter |
| Shape or size anomaly | Broken kernels, clusters, stalks | Shape or size analysis | Very small pieces may pass between ejector nozzles |
| Hidden mould or toxin risk | Mould-affected maize or peanut kernels | Multispectral (visible + NIR) with careful calibration | Results vary; verify with laboratory testing |
A useful research example comes from Kenya. Researchers calibrated a low-cost multispectral sorter at nine wavelengths from 470 to 1550 nm to detect maize kernels contaminated with aflatoxin and fumonisin. In calibration it identified contaminated kernels with 77% sensitivity and 83% specificity. Sorting 46 market samples gave an 83% mean reduction in each toxin, while rejecting 0 to 25% of sample mass (Stasiewicz et al., 2017). That was a research machine and study, not a guarantee for any commercial sorter. It shows why calibration on real contaminated material matters.
How are capacity, purity and product loss connected?
Short answer: A sorter has a limited number of ejector firings per second and a limited viewing area. Pushing more product through crowds the stream, so more defects slip past and more good pieces are blown out alongside rejects. Always ask for capacity at a stated purity and a stated good-product loss.
Define three numbers with your supplier before any trial:
- Feed rate in kg/h, with the defect level in the feed (for example, 2% discoloured kernels by count or weight).
- Accept purity: the defect level left in the accepted stream.
- Good-product loss: the share of good product that ends up in rejects.
Worked example (illustrative numbers, not a machine guarantee): feed 2,000 kg/h containing 2% defects. Defects = 2,000 × 0.02 = 40 kg/h. Good product = 1,960 kg/h. Suppose the trial shows the sorter removes 90% of defects and also rejects 1.5% of good product.
- Defects removed = 40 × 0.90 = 36 kg/h. Defects left in accepts = 4 kg/h.
- Good product lost to rejects = 1,960 × 0.015 = 29.4 kg/h.
- Accepted stream = (1,960 − 29.4) + 4 = 1,934.6 kg/h, with a defect level of 4 ÷ 1,934.6 ≈ 0.21%.
- Reject stream = 36 + 29.4 = 65.4 kg/h, of which about 45% is good product.
That last line is why many plants re-sort the reject stream. A second pass at a lower feed rate can recover much of the good product, while the accept stream keeps its purity.
What should a sorting trial include?
Short answer: Send enough of your real product, including typical and worst-case defect levels. Ask for results at your target feed rate, with accept and reject samples weighed and counted. For food-safety defects, have a laboratory test both streams. A short visual demonstration is not enough to sign a purchase order.
- Prepare a defect reference set: physical samples of each defect you want removed.
- Measure the feed: moisture, size range and defect level by count and weight.
- Run at least two feed rates, including your design rate.
- Weigh accepts and rejects. Hand-pick samples of each to count defects and good pieces.
- For mycotoxin concerns, send accepts and rejects for laboratory analysis. The Codex code of practice for peanuts states that sorting techniques should be checked by regular aflatoxin analysis of the sorted stream or finished product (Codex CXC 55-2004).
- Record settings so the result can be repeated on site.
Where should the sorter sit in the line?
Short answer: Put the optical sorter after cleaning and size grading, when dust, stones and size variation are already reduced. Uniform, clean, single-layer feed gives the cameras the best chance. For products that are blanched, roasted or split, a second sort after that step can catch defects that only show later.
The Codex peanut code supports this logic. It advises that sorting should be done before and after blanching and roasting, and that sorting machines should be adjusted regularly against standards. It also notes that colour sorting combined with blanching has been shown to reduce aflatoxin by as much as 90% in peanuts. Treat that as a published reference figure, not a plant guarantee.
Typical order for grains, pulses and nuts: pre-cleaning and destoning, size grading, optical sorting, metal detection, then packing. For spices such as chilli or turmeric, sorting usually comes before grinding, because defects cannot be removed once powdered. Our red chilli powder plant planning guide shows where sorting fits in a full spice line.
What utilities and site conditions matter?
Short answer: Optical sorters need clean, dry compressed air for the ejectors, a stable power supply, and a dust-controlled room. Oil or water in the compressed air damages valves and can contaminate food. Dust on viewing glass reduces accuracy, so plan cleaning and extraction from the start.
- Compressed air: ask the supplier for air consumption at your defect level. It rises as more pieces are rejected. Fit filtration and drying suitable for food contact.
- Dust: link the sorter to dust extraction and set a routine for cleaning viewing windows and lights.
- Temperature and vibration: follow the supplier’s limits. Mount the machine on a stable structure.
- Operator skill: someone must own the sort settings and recipes for each product.
What should you ask a supplier before buying?
- Capacity at stated accept purity and good-product loss, for your product and defect list.
- Sensor types, viewing (one or both sides) and minimum detectable defect size.
- Ejector count and spacing, and compressed air demand.
- Re-sort options and conveyor arrangement.
- Cleaning access and hygienic design of product-contact parts.
- Training, spares and remote support arrangements.
Frequently asked questions
Can an optical sorter remove all aflatoxin?
No. Sorting can reduce contamination where affected kernels look different, but some contaminated kernels look normal. Laboratory testing remains the only way to confirm compliance.
Is a belt or chute sorter better?
It depends on the product. Free-flowing, uniform grains often run on chutes. Fragile, irregular or sticky products may suit belts. Decide through trials on your material.
Do I still need manual inspection?
Often yes, at least for checking. Use hand-picked samples to confirm sorter performance each shift.
Related reading
- How to choose a grading machine for spices, grains, nuts and chips
- Choosing gyratory sifter screens and decks
- Spices processing plant and machinery
Talk to GTI about your project
Planning a sorting stage for grains, pulses, spices, seeds or nuts? Contact Genex Tech Industries LLP (GTI Group) to discuss your material, production capacity, product specifications and utility requirements. To help us respond usefully, please share your product or raw material, required capacity, feed and final moisture (state wet or dry basis), finished-product specification, available utilities, installation location and project timeline.
Email: mktg@foodtechprojects.com
Call or WhatsApp: +91-93300-77417 | +91-87774-75596 (WhatsApp)
Website: https://foodtechprojects.com/
For Industrial Dryers: https://gtidryers.com
This article gives general technical guidance. Figures marked as illustrative are examples, not plant guarantees. Equipment selection and performance for your material should be confirmed through trials and engineering assessment.
References
- Stasiewicz M.J. et al. (2017). Multi-spectral kernel sorting to reduce aflatoxins and fumonisins in Kenyan maize. Food Control 78:203-214 (CGIAR repository record). cgspace.cgiar.org
- Codex Alimentarius. Code of Practice for the Prevention and Reduction of Aflatoxin Contamination in Peanuts (CXC 55-2004). FAO/WHO. www.fao.org
