Waste characterisation: straight answers

It is the measurement of the composition of a waste stream using computer vision, artificial intelligence and sensors, continuously on a line or on demand on a skip or a bin. It complements manual characterisations, which only cover samples.

In its coverage. At Allier Tri, contractual characterisations cover less than 10 t a year, out of 20,000 t processed; the analyser continuously reads all the objects visible on the line. The two approaches are complementary: manual characterisations are used to calibrate and check the automatic measurement.

It can complement the measurement set-up and increase the frequency of observation. Whether a manual characterisation or a specific protocol must be kept depends on the intended use, the contractual commitments and the applicable requirements. Comparison with a reference remains essential.

It is built site by site. Each system is calibrated by comparing its reading with time-stamped manual characterisations, using the site’s grid. Accuracy is not the same for an object count as for a mass estimate, and results depend on the configuration of the line and on the incoming waste stream.

Depending on the solution, it can identify visible objects, count categories, estimate a surface area or a volume, or use a weighing. A mass estimate must be distinguished from a measured weight.

We set them out from the start. A camera does not measure moisture and only identifies visible objects. Without a sorting line, mass is estimated from density and the contents of closed bags are not analysed. In textile sorting, near-infrared analysis only sees the surface: complex blends, some dyes, black fibres and conveyor speed remain difficult cases.

No. It measures, alerts, and can control extraction equipment when such equipment is chosen. Operatives can then concentrate on the tasks where they are most useful: at Trigironde, automated extraction, still under study, would allow the sorters to refocus on fibre-based materials.

Installation takes a few hours: less than two hours on the Trigironde residue line. The real delay is calibration. It took two to three weeks at Trigironde, and two months at Allier Tri, where the classification grid was entirely reviewed with the operator.

No. At the start, we produce regular reports, help interpret trends and turn recurring drifts into simple alerts. Together with the site, we choose a few useful indicators, rather than multiplying dashboards.

Yes, from the start. At our reference sites, calibration was carried out with the operator. The data helps the operator understand the process, link a performance gap to the quality of incoming waste and know the limits of the line. It is a basis for dialogue, not a control tool.

No. You need to compare the full cost of the system with the service provided: installation, any subscription, sample preparation, checks, analysis time and use of the results. Targeted monitoring can be more relevant than a widespread deployment.

Arrangements vary depending on the solution and the assignment: services, equipment, licences or rental. The quote distinguishes these components. Your finance department determines how they are treated according to the contract and the applicable rules.

You remain the owner of your data. The rights of access, use, export and return are defined in the contract: we specify the roles of the supplier, the operator and the local authority, as well as the conditions for retention and reversibility. The data is returned to you as dashboards and exportable reports.

It documents the composition of waste streams continuously and over time: a factual basis for traceability, audits and dialogue with producer responsibility organisations. The stakes are rising with extended producer responsibility (EPR) for commercial and industrial packaging, in force since 1 January 2026, whose operational implementation was set for 1 January 2027.

With a thirty-minute conversation to scope your challenges and your waste streams. Depending on the case, we propose a time-limited audit, a feasibility study, an owner’s advisory assignment or a deployment directly on the useful measurement points.