Unmanned AI FFB Grading

From bunch inspection
to a connected grading record.

Unmanned AI FFB Grading brings the intake workflow together. Bunches are individualised and rotated for RGB, socket, spikelet and thermal inspection. The system supports grader acknowledgement, quality-based sorting, loose-fruit segregation, electronic grading chits, supplier records and weighbridge integration.

“World’s first palm oil mill to adopt AI” — The Star, 19 Nov 2024 · See the rig on real fruit

Book a site assessment Request the technical spec

A shipping container raised on a red steel plinth beside a mill building, carrying a mounted black signboard lettered UNMANNED AI FFB GRADING in yellow. A grey feeder housing joins its left end where the mill wall meets it, and a yellow access stair climbs to the right-hand end. A long covered conveyor gallery crosses the foreground, under a dark storm sky. — Installed unit **container-mounted, crane-set on plinth** Unit layout **twin inclined conveyors in · sealed grading box · sorted fruit out** Seen here **feeder housing at the mill wall · access stair**

The rig’s own output

This is the rig looking
at real fruit.

Not a diagram. One frame off a live belt, with the machine’s own masks burnt in. Everything coloured was drawn by the rig.

Overhead view of a load of fresh fruit bunches in a sectioned steel bed, seen from the grading rig’s own camera. Three translucent rectangular zones are drawn over regions of the load — two green, one yellow. Along the top edge of each zone a strip is blurred out: that is where the rig printed its grade label and confidence score. Pale lettering ending FFB-GRADING, the system’s own on-screen title, runs across the lower right of the frame and passes under one of the green zones. — **01–03** The three regions the grading system localised in this frame. Each bracket sits on the position listed below.

**FILL** Inside a bracket the rig’s own segmentation fill runs bright; outside it is dimmed.On paper the frame prints undimmed; the brackets, labels and coordinates carry over.

**RGB** The colour channel is one of four. Socket, spikelet and thermal read the same bunch on the tiers that follow.

Source file
assets/img/vision-overlay-redacted.jpg

Frame
1280 × 720 · grading rig, FFB conveyor

Zone 01
x549 y352 → x744 y460

Zone 02
x704 y480 → x921 y629

Zone 03
x450 y563 → x663 y706

Overlay drawn by
the grading rig, at grading time

The machine drew these masks at grading time.

The coordinates are good to about **±10 pixels** and record **where the machine drew**.

A scuffed steel reject bin at a mill, part-filled with underripe dark oil palm bunches, standing on wet concrete beside a drain grate. — Where the load comes from

A mill’s intake is other people’s fruit. Every load is a supplier who will ask how the grade was reached.

A cut oil palm fruit bunch lying on wet ground, its fruitlets ranging from purple-black through deep red to orange, the pale fibrous cut face of the stalk at left and several loose fruitlets fallen in the mud beside it. — Where it sits

Beside the ramp, on its own plinth, wired into the weighbridge.

How it works

Fresh fruit bunches moving along a steel conveyor inside a mill under bright industrial lighting, loose fruitlets scattered across the belt. — What is being read

Ripeness is not one signal: fruitlet colour, how many detached, the sockets they left, the fibre between.

Four channels read the
same bunch.

One bunch at a time, turned so every side is seen. Four tiers read it, each asking a different question.

TIER 0

Individualizer & rotator

Separates the bunch and turns it through 360°, so the cameras see every face, not whichever landed upward.

TIER 1

RGB colour

Colour for ripeness — the signal a grader uses, but at a fixed distance under a fixed light, on every bunch instead of a sample.

TIER 2

Socket analysis

The sockets left by fruit that has already dropped. Colour can be argued with. Sockets cannot.

TIER 3

Spikelet detection

Spikelet detail shows freshness — how long ago it was cut. A different question from how ripe it is.

TIER 4

Thermal imaging

Heat, not surface. A bunch cannot be dressed up to fool a thermal camera.

Specification

Form factor: Container-mounted, crane-set on a plinth
Inspection channels: 4 — RGB · socket · spikelet · thermal
Bunch presentation: Individualizer & rotator, 360°
Sorting: Automatic, on the AI result
Loose fruit: Extracted separately
Rejects: Returned to the supplier
Weighbridge: Integrated
Grading chit: Issued digitally at the point of grading
Operators in the loop: None for the grade itself; acknowledgement from the grading cabin
Throughput: Ask, and we send it in writing with its duty cycle and bunch-size basis

Lorry to ramp

Loose palm fruitlets scattered across a wet concrete mill floor beside a drain channel, a worn yellow bollard and a hose at the edge of the frame. — On the belt

The rig does not stop the line. Bunches arrive at the ramp’s own rate, and grading keeps up or it is a laboratory, not a mill.

What happens to a load.

FFB is unloaded by the supplier.
The auto feeder brings the fruit into the grading box.
Inside the grading box, each bunch is individualised and rotated.
Four-tier inspection: RGB colour, socket, spikelet, thermal.
Automatic sorting on the AI result.
Loose fruit is extracted separately.
Ripe fruit goes to the loading ramp.
Rejected fruit is returned to the supplier.

Two men at the tailgate of a loaded fruit lorry on a mill ramp, looking together at a small rugged tablet held between them, loose fruitlets scattered across the ground and the mill’s inclined conveyor behind them in low afternoon light. — At the tailgate

The result is read where the load is standing. What used to be settled at the office days later is settled at the ramp.

A tipping trailer discharging a load of fresh fruit bunches into a mill receiving bay, the raised steel body streaked with grease and rust in hard daylight. — At the gate

Rejects go back on the lorry they came in on, and the chit is issued at the ramp. A supplier leaves with the result, not a promise of one.

What the mill sees

The grade stops being
an argument.

One record, made at the moment of weighing, is harder to argue with than a judgement made in the heat.

A weathered hand passing a small printed paper docket through the open window of a lorry cab to the driver, whose hand reaches out to take it. Dusty windscreen, mill sheds out of focus behind. No face is visible. — The console

Grading moves off the ramp, out of the heat and dust, onto a screen with a record behind it.

Control room

The line is watched from a room, not the ramp. The cabin acknowledges; it does not assign the grade.

Long-stalk detection

Long stalk gets its own percentage, alongside freshness. The deduction suppliers dispute most becomes a measurement, not a judgement.

Supplier statistics and historical pattern

Quality accumulates per supplier — a trend you can show them, and a basis for pricing that survives a conversation.

Weighbridge integration

Grade and weight are captured against the same vehicle in one transaction, so they cannot drift apart.

Instant digital grading chit

The e-grading chit is issued at the point of grading. Nobody types it in afterwards.

A steel sorting table inside a mill with oil palm fruit separated into two rough groups by ripeness, the riper fruitlets deep orange-red and the underripe ones darker. — Where the two numbers meet

Suppliers argue about the deduction, not the weight. One ticket carries both, so there is one document to talk about.

What the rig grades

An industrial machine-vision camera and LED bar mounted on a steel frame above a mill conveyor, cables routed along the frame, haze in the air catching the light. — The instrument

A camera, a light bar, a frame, and a mill's own dust and vibration. Accuracy belongs to that whole assembly, not the model inside it.

Every bunch read on four channels.
Nobody types the grade in.

Every bunch is read on four channels — RGB colour, socket, spikelet, thermal. The machine issues the grade; nobody types it in.

The system grades “with an accuracy of 98%” — as published on airei.com.my/ai-ffbgrading.

Ask us for the basis and we send it in writing, in a day.

On request

The accuracy figure

Ask, and we send it: In writing, in a day, with its basis attached.
What comes with it: The grading method it was measured against, how many bunches, over what period, who ran the comparison, and which model build produced it.

Request the figure and its basis →

A long LED illumination bar mounted above a mill conveyor, its diffuser scratched and dust-flecked, throwing hard even light onto oil palm fruit below. — The same light, every bunch

The bar throws hard even light onto the belt, so every bunch is read under the same conditions as the last.

Next step

Two ways in, and neither
of them is a brochure.

A site assessment tells you whether the rig fits your ramp, your weighbridge and your queue.

The flank of a white shipping-container unit standing on a red steel plinth beside a palm oil mill, a cable gland plate and a vent on its side. — What a visit settles

Ramp geometry, weighbridge wiring, queue length, and what the retrofit has to work around.