Green waste: artificial intelligence detects foreign matter

Foreign material in green waste is something that many fermentation plants and composting sites have to contend with. In a pilot phase at Axpo's Chavornay fermentation plant, artificial intelligence is helping to improve the quality of the delivered material.

Every day, an average of 10 lorries unload green waste at Axpo's Chavornay fermentation plant, which amounts to around 25,000 tonnes per year. Unfortunately, however, these deliveries also contain materials that do not belong there: Metal, glass, waste paper, aluminium foil, plastic and much more (see info box). Why is this a problem? ‘Such foreign materials cannot be processed in the fermentation process and consequently impair the quality of our soil,’ explains Denis Crausaz, head of the Axpo fermentation plant in Chavornay. This is because the fermentation of green waste not only produces valuable energy, but also nutrient-rich natural fertiliser that is spread on the regional fields. Cleaner waste also increases the efficiency of CO₂-neutral biogas production.

All fermentation plants and composting sites in Switzerland are familiar with the problem of foreign matter in green waste. There are technical facilities or checks by employees to sort out foreign matter. However, the most efficient approach would be to tackle the problem at its roots. In other words, to ensure that people only put what is really intended for the green waste collection centre.

This is exactly where the pilot project in Chavornay comes in. ‘We use cameras in the refuse lorries and in the fermentation plant,’ says Denis Crausaz. Behind the pilot project is the Vaud-based IT company Cortexia, which specialises in technical solutions for environmental and waste issues. 

And this is how it works: The basis is a special rubbish truck equipped with a camera and two powerful spotlights. It uses these to take pictures of the material in the green waste bins. These images are analysed in real time by an external computer on the vehicle, the so-called ‘Cortexia box’. The artificial intelligence can identify a range of foreign substances and sends this data, together with the GPS coordinates of the container, to the Chavornay fermentation plant. After its journey, the vehicle reverses to the fermentation plant's pit. There, an additional camera monitors and analyses the delivered garden and food waste once again. 

A lot of work went into this solution. The algorithm had to learn to recognise foreign matter from the large variety of garden and food waste. It took around a year to reliably recognise the long list of unwanted materials. The algorithm also had to be able to differentiate between compostable and non-biodegradable plastic bags. 

Cortexia's solution provides an accurate picture of which localities and neighbourhoods deliver the most foreign materials. The aim of the project is to create a precise colour-coded map of the quality of the biowaste collected in the northern part of the Vaud region. Depending on the proportion of unwanted materials, there will be green, yellow or red zones. 

Thanks to this mapping, the regional waste association Strid will be able to carry out a targeted prevention campaign. In other words, where the impact of such a campaign can be greatest. The first wave has been running since autumn 2024 in Yverdon-les-Bains, where the map shows orange and red zones. Will it work? The Axpo fermentation plant will soon find out - even in real time thanks to artificial intelligence.