Produce Grading and Sorting

• Packhouse Operations Manager
• FPO Procurement Head
• Quality Assurance Lead, Agri Sourcing
• Head of Supply, Agri Marketplace

At a collection centre or packhouse, grading decides what the farmer is paid and what the buyer accepts. It is often done by eye on a sample, by a grader who may be tired by the afternoon, and disputes over grade are settled by argument rather than evidence.

We build a vision-based grading station for sample trays or conveyor lines. A fixed camera under controlled lighting captures each sample or piece, and models measure size, shape, colour and surface defects such as bruises, cracks, rot and pest damage. For grain, the system counts and classifies kernels on a tray image, including broken, damaged and foreign matter. Grade rules are configured per commodity and buyer specification. Each lot gets a grade, a defect breakdown and stored images, linked to the farmer and weighbridge record. Borderline lots and disputes go to a QC assayer, who views the same images and can override the grade with a reason.

• Size, shape and colour measurement for fruit and vegetables
• Surface defect detection for bruising, cracks, rot and pest damage
• Kernel-level grain sample analysis for broken, damaged and foreign matter
• Configurable grade rules per commodity and buyer specification
• Lot-level grade record with stored images
• QC override with mandatory reason capture

1. 01Capture the sampleThe operator places a sample on the tray or lets produce pass the conveyor camera, and scans the lot or farmer token.
2. 02Measure and detect defectsModels measure each item or kernel and mark visible defects and foreign matter.
3. 03Apply grade rulesMeasurements are compared with the commodity and buyer specification to produce a grade and defect breakdown.
4. 04Check borderline lotsLots near a grade boundary, or with poor image quality, are flagged instead of being auto-graded.
5. 05QC assayer reviewThe assayer reviews flagged lots and any farmer or buyer dispute using the stored images, and confirms or overrides the grade.
6. 06Publish the resultThe grade and images are attached to the lot in the procurement system and shared with the farmer and buyer.

1. Capture station
   • Fixed camera with diffuse lighting enclosure
   • Conveyor or tray trigger
   • Lot token or QR scanner
2. Vision models
   • Segmentation per fruit or kernel
   • Defect detector
   • Colour and size measurement
3. Grading logic
   • Commodity and buyer grade rule engine
   • Borderline detection
4. Review and records
   • QC assayer review screen
   • Image archive per lot
   • Override log with reasons
5. Integration
   • Procurement or FPO ERP adapter
   • Weighbridge data link
   • Farmer receipt by SMS or WhatsApp

QC assayers review every borderline lot and every dispute raised by a farmer or buyer, using the same stored images. Overrides require a reason, which is used to recalibrate grade rules and models.

• Procurement or FPO ERP
• Weighbridge software
• Farmer payment system
• WhatsApp Business API or SMS gateway for farmer receipts
• Buyer portal or dispatch system

• Lighting and camera calibration check at the start of each shift
• Borderline lots never auto-graded
• All grade overrides logged with user and reason
• Periodic blind re-grading of sampled lots by QC staff
• Separate model versions per commodity, released only after QC sign-off

• Where produce is graded under AGMARK, grade rules can be configured to the published grade standards, with the authorised grader keeping final responsibility.
• Images and farmer identifiers are stored with access controls and retention periods agreed with you.

• Agriculture & Agritech

  Crop Disease Detection from Farmer Photos

  Suggest likely diseases, pests and nutrient problems from farmer photos sent on WhatsApp or an app, reply in the farmer's language, and route uncertain or dose questions to an agronomist.

• Agriculture & Agritech

  Warehouse Receipt and Quality Certificate Extraction

  Extract and cross-check warehouse receipts, e-NWRs, quality certificates and stack details for commodity finance, so credit and collateral teams catch mismatches before disbursal.

• Agriculture & Agritech

  Mandi Price and Procurement Voice Updates

  Outbound and inbound voice updates on nearby mandi prices, procurement windows and pickup slots in the farmer's language, with every price dated and sourced, and offers approved by procurement.

• Manufacturing

  Visual Quality Inspection on Production Lines

  Cameras and vision models at the inspection station flag scratches, dents, missing features and print errors, divert clear rejects and hold borderline parts for an inspector.