TL;DR
- An industrial camera films every lift, an AI computer on the truck recognises the waste and an online platform turns the results into reports and alerts.
- With D4T from Rematics, this happens in five steps: record, recognise, process, create insight and take action.
- The analysis runs at 40 images per second. That way the system sees the entire waste dump, not just a snapshot.
AI waste recognition on a refuse truck works with three components: an industrial camera, lighting and an AI computer that does its calculations on the vehicle itself. During every lift, that computer analyses what falls into the hopper. The results end up in an online platform per collection, customer or location.
Below you can read how this works step by step, using D4T (Daisy4Truck) from Rematics as an example. More than 70 refuse trucks run with this technology today.
Step 1: record
The camera captures the entire lifting process, from the moment the container tips until it is empty. Dedicated lighting provides usable images, even on rounds that start at 3 or 4 a.m. The camera sits in a housing that withstands dust, vibration and rain.
Step 2: recognise
Computer vision analyses the images while the waste falls. D4T does this at 40 images per second and recognises more than 40 waste fractions, such as paper, cardboard, PMD, film, metal, wood and textiles. The system also spots non-conformities and hazardous materials such as gas cylinders, nitrous oxide canisters and batteries.
Why so many images? Waste falls in layers. What lies on top hides what is underneath. Taking just a few photos means missing part of the contents. "The more images we analyse, the more reliable the data," says Pieter Schiepers of Rematics.
Step 3: process on the vehicle
The analysis takes place on the AI computer in the truck, at the edge. So the truck does not need to send hours of video. Every detection immediately gets a time and a location and is linked to the collection, the customer or the address.
If the system detects a hazardous item, the driver immediately hears an audio signal in the cab.
Step 4: create insight
Via 4G, 5G or wifi, the results go to the Daisy Portal, the online platform from Rematics. There you find dashboards, reports, maps with hotspots and the images of every detection. You look at three levels:
- during the collection, with real-time alerts;
- after each round, with an overview of collections, materials and risks;
- over months, with trends, hotspots and the effect of campaigns.
Step 5: take action
Data changes nothing as long as nobody uses it. That is why the last step is about follow-up. A private waste collector shows a business customer, with images, what was in their residual waste. An inter-municipal waste authority sees which neighbourhoods need extra explanation.
Milieu Service Nederland uses the insights in advisory conversations about source separation and collection frequency. That way a camera image becomes a concrete conversation with the customer.
What you need to get started
Installation takes less than 4 hours per truck on average. D4T works with wheelie bins, waste bags and large commercial containers. In the frequently asked questions you can read how the system copes with rain, darkness and dirty bags.
Want to see the five steps in action? See how D4T works or request a demo and find out what your fleet can already measure today.