customer story
Waste sorting
Machine vision identifies and classifies heterogeneous flows in real time by type, shape or color, to guide a sorting robot with a precision that conventional detection systems cannot achieve.
Contact usWhat is machine vision used for in industrial waste sorting?
Machine vision watches the belt continuously and tells you what is passing: material types, proportions, piece size, presence of contaminants. Its first use is therefore to measure the real quality of a stream, at the input as well as the output of each sorting stage, where all you had before were samples taken with a shovel. On that basis, it then guides robotic sorting and flags drifts before an entire batch is downgraded.
- A stream characterized continuously, not sampled
- Pieces recognized by material and texture, not by shape
- A purity measurement available at every sorting stage
- Models that keep up with the drift of the feedstock
What a vision gantry changes on a sorting line
100 %
of the stream observed
Manual characterization samples a few kilos; the gantry looks at everything that runs beneath it.
24/7
with no lapse in attention
The composition of the stream is tracked continuously, including at night and during shift changes.
2.5D
for gripping
To ensure picks at varying heights, a Z sensor is added to the 2D camera.
Industry challenges
What makes sorting harder than quality control
A stream you do not choose
The feedstock depends on suppliers, the season and the market. No production plan says in advance what will arrive on the belt tomorrow.
Unrecognizable pieces
Shredded, bent, crushed, burnt or soiled: the shape no longer says anything about the original product. Recognition relies on material, texture and color.
A dirty belt
Dust, mud, moisture, vibration, pieces that overlap and hide each other: a usable image has to be obtained despite all of that, not in ideal conditions.
A drifting composition
What was right six months ago no longer is. A frozen model slowly falls behind, with nothing to signal it if nobody measures it.
Knowing your output
Purity rate, losses in the reject stream, actual efficiency of each separator: all figures that drive the value of the bales and that sampling estimates poorly.
Huge volumes of images
A belt produces millions of images. Storing, annotating and using them takes proper tooling, not a spreadsheet.
On the line
A gantry mounted above the belt
The Psycle vision gantry is installed above an existing conveyor, without touching the process. Camera, lighting and housing are sized for the dust, moisture and vibration of the line, and to produce a usable image on a belt that is never clean.
It continuously characterizes what passes beneath it, and is placed where the question arises: at the input to know what you are receiving, after a separator to measure what it lets through, before baling to qualify the bale you sell.
Under the hood
Large image databases rather than a catalog of rules
No rule written in advance describes a shredded piece. The model therefore learns from images of your stream, including the cases your operators debate, and keeps learning as the feedstock changes.
That means being able to live with large volumes: collecting without saturating the line, finding the images that matter, annotating quickly, retraining and comparing versions. That is exactly what PAQ : The Psycle quality monitoring and fleet management platform, available in both cloud and on-premises versions. does, and it is what lets your teams take over the model without going back through a service provider.
customer story
3D bin picking
3D localization of loose parts allows the robot to calculate the optimal pick without a dedicated gripper or organized feeding. At Psycle, vision models adapt to the diversity of shapes and materials.
Contact usMethod
How a project unfolds on your sorting line
.01
Feasibility
Your feedstock, not someone else's
The study starts from images taken on your belt. It shows which materials can be separated, at what particle size, and which cannot.
.02
Trial gantry
Measure before sorting
A first gantry characterizes the stream and provides the training images. Many projects usefully stop at this stage.
.03
Production rollout
Sorting and supervision
Guidance of the ejectors or the sorting robot, purity indicators fed back to your operations tools.
.04
Autonomy
The model keeps up with the feedstock
New material, new supplier, seasonal drift: your teams retrain and redeploy without us.
Where the gantry goes on the line
Each position answers a different question, and it is the question that decides the location, not the other way around.
| Where the gantry goes | What it provides | |
|---|---|---|
| Line input | On the infeed conveyor | Composition of the incoming stream, by supplier and by truck |
| After the shredder | At the shredder outfeed | Particle size and materials present after size reduction |
| After a separator | On the sorted output and on the reject stream | Actual efficiency of the separator, and the losses it generates |
| Before baling | Before the baler | Purity rate of the batch, measured before it is sold |
| On robotic sorting | Above the picking station | Identification and position of the pieces to remove from the stream |
| On the final reject stream | Before landfill or energy recovery | Lost value, and the case for an additional sorting stage |
They sort with Psycle
All testimonials
With Psycle, we have reached a new level.
Charles Lestoquoy – Director
Ascodero (groupe Siléane)
Recycling and recovery companies
Frequently asked questions
Full FAQIndustrial waste sortingCan a vision gantry be installed on an existing sorting line?
Yes. The gantry is installed above an existing conveyor, without modifying the process. The camera, lighting and housing are designed for the dust, humidity and vibrations of the line.
Industrial waste sortingHow can shredded, bent or soiled items be recognized?
Their shape no longer says anything about the original product, so recognition relies on material, texture and color. The model is trained on images from your own flow, including the borderline cases your operators debate.
Industrial waste sortingWhere should the gantry be placed on the line?
The location depends on the question you are asking:
- at the infeed, to know what you are receiving;
- after a separator, to measure what it lets through;
- before baling, to qualify the bale.
Several gantries make it possible to monitor each sorting stage.
Industrial waste sortingDoes the model remain reliable when the feedstock changes?
A frozen model gradually loses accuracy. That is why the accumulated images are used to retrain it as the flow evolves. With PAQ, your teams collect, annotate, retrain and compare versions themselves, for example when a new material or a new supplier arrives.
Industrial waste sortingCan vision replace manual characterization by sampling?
It largely complements it. Where sampling takes a few kilos, the gantry analyzes everything that passes, day and night, including during shift changes. It continuously tracks the composition of the flow, contaminants and particle size.
Industrial waste sortingDo you have to go as far as robotic sorting to benefit from vision?
No. Many projects usefully stop at the characterization gantry: you finally know the real purity of your output, the losses in the reject stream and the efficiency of each separator. Controlling ejectors or a robot can come later, based on the images already collected.