Product · Matvision Sense
Know what is in your stream, piece by piece,
before you decide.
The sensing and decision core of every Matvision machine, available on its own to characterise a stream or a lot: composition by material family and, with LIBS, by alloy grade.
Applications
Built for every stream.
Sense runs on any stream a Refine or a Grade can sort, and on the ones you are still deciding about.
E40 steel scrap
How much copper and attached material the magnet carried along, measured on your own load.
Zorba and Zurik
Composition by family, and by alloy with LIBS, before you size a sorting line.
WEEE and batteries
What came in and what went out: a traceable identity for each lot.
How it works
One piece,
five measurements.
- 01
Material presented
The belt sets speed, regularity and layer thickness. Without that controlled presentation, no measurement is usable.
- 02
Height profile
A line laser projects onto the piece and an inclined camera watches that line from the side. The way it deforms across the surface gives the relief.
- 03
Surface and colour
The colour camera and its own lighting read what the surface shows, and only that.
- 04
What the surface hides
The source fires through the piece to a detector set inside the conveyor. What the surface conceals, the transmission reveals.
- 05
Which elements
A focused laser shot vaporises a point of material. The plasma emits, and the spectrometer splits that light into lines that name the elements.
- 06
Spectral bands
A surface does not return every wavelength the same way. A line camera splits the light it sends back into bands far finer than colour, and that signature tells apart materials a colour image reads as identical.
- 07
Five measurements, one piece
Each sensor returns something the others cannot. They describe the same piece, at the same instant. That simultaneity is the whole argument.
- 08
Into the model
The measurements go in together. It is their combination that names the material, never one of them on its own.
- 09
The verdict
Quality, size, production. The report carries the composition and the software version, model and thresholds that produced it.
The problem it solves
Every sorting decision starts with four numbers you rarely have.
Annual tonnage, current composition, current resale price, target price. Those four numbers turn a technical discussion into an economic one, and they decide whether an extra sorting stage pays back in one year or in three. Most plants know the tonnage and the prices. The composition is usually an estimate, sampled by hand, on a stream that changes every week.
Matvision™ Sense measures it. Each piece that passes under the sensors is identified and counted; the stream is described as it really is, not as the last manual sampling said it was.
Why several sensors
A camera sees a surface.
It does not see a piece.
A colour camera reads the light that comes back off a surface: colour, texture, the shape of the outline. That is real information, and on some streams it is enough. It is also everything a camera can give you. It cannot tell you how thick a piece is, what sits under the paint, or which elements it is made of.
Matvision Sense measures those too, on the same piece and at the same instant. The relief gives it a volume rather than an outline, transmission looks through it, and the elemental measurement names what it is made of. That is the difference between recognising a piece and characterising it.
Volume, not outline
A flat sliver and a solid block can look alike from above. They do not weigh the same and they do not have the same effect on a lot. Measuring the relief is what turns a contaminant that has been detected into a contaminant that has been quantified.
Under the surface
Paint, coating and dirt change what a camera sees. Transmission and elemental measurement do not stop at the surface, so a piece is judged on what it is rather than on how it looks that day.
Several readings of one piece
Combining geometry, colour and X-ray transmission helps distinguish pieces that look similar on the surface. Several measurements of the same piece, taken at the same instant, either agree or they do not, and that is a different kind of ground to stand on.
Which of them you need is a question about your stream, not about our catalogue. The combination is chosen on your own material, during the trial: the sensors that actually separate the classes you care about, arranged so that each one covers what the others cannot. A stream where colour and relief settle the question does not need a spectrometer. One where two alloys look alike does. The characterisation is configured for the stream in front of it, and that is what makes it both reliable and proportionate.
Characterisation
When data becomes an asset.
A heap is only a heap until each piece in it has been identified. Then it is a composition you can price, sell and plan with.
What you do with the numbers
Quality control on both sides of the gate.
A composition measured piece by piece is not only a technical result. It is something you can put in front of a supplier or a buyer, and it works in both directions: on the material you take in, and on the material you send out.
On what you buy
Placed at the input, Matvision Sense describes a delivery as it actually arrived. When a lot does not match what was sold to you, the discussion rests on a measurement taken across the delivery rather than on a shovel taken off the top of the pile.
On what you sell
Placed at the output, Matvision Sense describes the lot before it leaves: composition in pieces, weight and percentage, size distribution, and the software version, model and thresholds behind the figures. The report goes out with the lot.
While the line runs
The composition is followed over time, so a stream that drifts away from what the models were trained on shows up while it is happening, not at the next manual sampling.
All of it lands in a dashboard reached from a browser, behind a login: the production that is running now, and the history of the ones before it, from the control room or from anywhere else.
And the numbers do not have to stay on a screen. The same output can hold a threshold and raise an alarm, drive a bulk decision on a whole batch, or become the instruction of an ejection stage (air ejection or a robot) the day you decide to act on it rather than only read it.
Application cases
Where a measurement changes the decision.
Characterise a Zorba, Zurik or E40 stream before an investment
Composition by material family: copper, zinc, lead, aluminium, stainless, nickel-plated. Share of contaminants such as printed circuit boards or lead pieces. Size distribution. With LIBS, the split inside a family: stainless grades, cast against wrought aluminium, wrought series. The output is the input of your return-on-investment calculation and of the configuration of a Refine or a Grade system.
Sensors: 3D, colour, XRT, LIBS optional
Run a trial before any sorting instruction is fixed
This is how every Matvision project starts, and it happens before anything is decided. You send two big bags: one pre-sorted by your team into the classes you care about, in the real proportions of your stream, and one mixed and untouched. The first trains the models. The second is processed blind, at target speed and feeding density, to measure recovery per class, purity of each fraction and achievable throughput. You leave with numbers measured on your own material, and with the sensor stack that is economically justified rather than the most complex one.
Give a lot a traceable identity
The same report is what a digital passport needs: a measured composition, and the software version, model and thresholds behind it, attached to a lot that has left. It is the basis of the passport on the Spectra roadmap.
Pilot and laboratory characterisation
For a research partner or a plant preparing a new stream, a Matvision Sense unit with one or two sensors and a low throughput is the smallest way to start collecting real data.
Configuration
The right sensors for your material.
A Sense line is built from one sensor or from several. No configuration carries them all: the combination follows your material and the separation you are after. Sense measures; it carries no robots by default.
RGB
RGB colour camera
What it measures
The colour, texture and outline of the visible surface.
Why it matters
Tells apart materials that differ by their surface: copper, brass and grey metals, painted or coated pieces.
3D
3D profilometer
What it measures
The height profile of each piece: shape, thickness and volume.
Why it matters
Describes geometry and tells a robot where to grip. It supports identification; it does not name a material on its own.
XRT
X-ray transmission
What it measures
How much X-ray passes through the piece, which follows the density of the material and what it is made of.
Why it matters
Looks through the piece, whatever its surface: a copper winding inside a steel casing, a heavy metal among light ones.
XRF
X-ray fluorescence
What it measures
The X-rays that the elements at the surface of the piece emit in return.
Why it matters
Reads which elements are present. Where XRT looks through a piece, XRF reads its surface, so coatings and dirt count.
VNIR
VNIR spectral imaging
What it measures
The light reflected across the visible and near-infrared range, band by band.
Why it matters
Adds spectral detail a colour camera cannot resolve. Whether it helps on a given material is established by testing.
SWIR
SWIR spectral imaging
What it measures
The light reflected in the short-wave infrared range, band by band.
Why it matters
Considered for separations the other sensors do not settle. Its relevance is established case by case, on the material.
MWIR
MWIR infrared imaging
What it measures
Radiation in the mid-wave infrared range.
Why it matters
Considered for separations the other sensors do not settle. Its relevance is established case by case, on the material.
LIBS, for elemental composition and alloy recognition, completes the stack where the separation calls for it.
Sensor selection depends on your material and separation goals. The appropriate combination is validated through testing on your own material.
What we measure, what we promise
A sample.
A separation.
A measured result.
Matvision Sense produces measurements, not commitments. Detection performance depends on the stream, on the sensors selected and on the trade-off between purity and recovery, which is why every figure we quote comes from a performance study on the material concerned.