Turning the Potato Over So the Camera Can See All of It Grading Bulbs by Appearance

This application covers a sorting device for agricultural products in bulk flow, principally bulbs such as potatoes, carrots and onions, which combines machine vision grading with a mechanism that rolls each product over so its whole surface can be inspected. The applicants are French: Ets Dubrulle, a potato handling equipment maker, and Bilberry SAS, an agricultural computer vision company.

The Problem

Sorting machines are already standard in potato handling, but they do a narrower job than their appearance suggests. The established technology uses infrared cameras or ultraviolet illumination to distinguish clods of earth and stones from the crop, and a finger separation mechanism then diverts the foreign bodies to waste. That works well and has been in the field for years.

What it cannot do is separate the crop from itself. Under infrared or ultraviolet, one potato looks much like another. The machine can tell a potato from a stone but not a good potato from a green, damaged, misshapen or diseased one. Grading by quality therefore still happens by hand, with people standing at a belt watching produce go past, which is slow, expensive and increasingly difficult to staff.

That leaves a gap in an otherwise automated line. The machinery handles the easy discrimination and leaves the commercially important one to human eyes.

What This Invention Does

The device grades on visual appearance rather than material type. It has an image acquisition unit recording images of the bulk flow, an image processing unit that identifies each individual product and analyses its appearance from those images, a decision unit that determines which products should be separated and issues a control signal, a separation unit that acts on that signal, and a conveying unit that carries the bulk flow from imaging to separation.

The distinguishing element is a means for turning over the agricultural products on themselves, mounted in the flow path. A potato photographed from above shows one side. Defects, greening and damage do not politely arrange themselves upward. Without a way to see the other face, appearance based grading either misses a large share of defects or requires a far more elaborate multi-camera arrangement. Rolling the product between imaging stages lets a straightforward camera setup see all of it.

Because grading is by appearance rather than by a fixed physical property, the selection criteria are adaptable. The same machine can be set for different crops, different customers and different quality specifications by changing what the processing and decision units are looking for.

Key Features

• Image acquisition of the bulk flow. Cameras record the flow of product rather than singulated items on a lane.
• Per product identification. The processing unit segments the images to identify each individual product within the bulk flow and analyse its visual appearance.
• Decision unit with control signal. A separate decision stage converts appearance analysis into a control signal for the separator.
• Turning means. A mechanism rolls the products over on themselves so surfaces hidden in the first view can be imaged.
• Separation unit driven by the signal. Products flagged by the decision unit are diverted out of the flow.
• Adaptable selection criteria. Grading on appearance allows precise and adjustable selection across crops and specifications.

Who Is Behind It

Ets Dubrulle trades as DOWNS, a French manufacturer of potato reception, handling and storage equipment whose lineage in the sector runs back to the nineteenth century and which acquired the DOWNS name in 2006. It launched an optical grader for unwashed potatoes under the CropVision name in 2021.

Bilberry was founded in Paris in 2016 by engineering graduates and built its reputation on real time weed recognition for selective spraying, cutting herbicide use dramatically by treating only the weeds. It was acquired by Trimble in 2022 and now sits within PTx Trimble. The named inventors are Damien Dubrulle, Hugo Serrat, Thomas Deschodt, Guillaume Jourdain and Duc-Anh Luu. The application is a divisional of Australian application 2020335104.

Why It Matters

The pairing tells you something about where agricultural machinery is going. An established equipment manufacturer that knows conveyors, throughput and the realities of a packing shed has teamed with a computer vision company whose expertise came from a completely different application. Bilberry learned to recognise weeds among crops at speed from a moving sprayer, and recognising a blemished potato among sound ones is the same class of problem.

Labour is the immediate driver. Manual grading lines are among the harder agricultural jobs to fill, and the work is repetitive enough that consistency degrades over a shift in ways a camera’s does not. There is also a waste argument: more precise grading means less good produce discarded with the bad, which matters when a meaningful fraction of horticultural output never reaches a shelf.

The turning mechanism is the part worth noticing. It is a mechanical answer to a perception problem, and it is the kind of unglamorous detail that determines whether a vision system works in a shed rather than in a demonstration.

Related Concepts

• Computer vision – the image analysis field behind appearance based grading.
• Optical sorting – the established machinery category this device extends.
• Precision agriculture – the wider movement towards per item and per plant decisions.
• Potato – the principal crop the machine is aimed at.
• Food loss and waste – the problem that better grading accuracy helps address.
• Image segmentation – the technique needed to pick individual products out of a bulk flow.

---

AU 2026202069 was published in the Australian Official Journal of Patents on 9 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.