2025-12-09 –, Auditorium
At waste transfer stations for source separated packaging waste incoming waste trucks are visually inspected on objects that could disturb the sorting and recycling of the truck load. This is a manual procedure and in case the number of disturbing items is too high, the part of the truck load needs to be removed. Currently, 8.5 % of the truck loads is rejected. This leads to loss of valuable plastics for recycling. We have investigated the automation of this inspection using cameras and vision foundation models. Inhouse, we developed a data pipeline where waste items are first detected, then segmented and eventually classified whether they belong to this waste stream using anomaly detection. The accepted material continues to a plastic recovery facility. This approach has led to a proof-of-principle with the potential to be implemented as a pilot-scale at a waste transfer station. The project is part of the research program ‘MultiPurpose Plastic Sorting’ subsidised by TKI Energy & Industry.
After waste collection of lightweight packaging waste, in the Netherlands known as PMD (plastic-, metaal- en drankkartons), the incoming material is visually inspected by an operator to extract contaminants that disturb the sorting and recycling. Typical contaminants are non-packaging material and large rigids and foils.
At the applied research institute NTCP, we investigated how to automate the detection process in order to have an objective method to inspect a PMD waste stream. This improves consistency and thereby quality and reduces required time needed for a thorough assessment. We decided to use cameras in a top view of the pile of items, as the inspection is performed. For every item, we first need to locate it in the image and segment the pixels of it. We built a data pipeline using vision foundation models.
To determine whether an item is considered a contaminant or PMD packaging, we found that the class of contaminants had a too broad variety to build a classification network. Therefore, we chose to explore the feasibility of anomaly detection models. Several dedicated anomaly detection models were trained. Finally, we learned that using a combination of foundation models for both visual as well textual features, resulted in a suitable distribution to distinguish contaminants from PMD packaging.
With a prototype set-up, we have showcased a proof-of-principle of a well working and promising automatic detection system on actual waste, given the variety in both PMD and contaminants. images, given the variety in both valid waste and unwanted items.
Tom Koopen is a highly experienced professional in computer vision and deep learning. He holds a Master’s Degree in Applied Physics from the University of Twente, specializing in optical measurement systems.
With over 25 years of experience in computer vision, Tom has worked with various companies in the Netherlands. Since 2013 he is an entrepeneur at “de tijdelijke expert”, assisting customers with the application of computer vision technology, focusing on measurements, identification, and sorting of products. Recently he founded “textilemining.eco” to build innovative machines for textile recycling.
He has designed lighting systems, selected and optimized cameras, and written software for thousands of hours. Some of his notable projects include inspecting plastic crates for contamination, improving the sorting of plastics, metals and flower bulbs. He developed a 3D scanner to recognize roof tiles for Luijtgaarden B.V. and measured colors during high-speed printing processes at QI Press Controls. Oh, and don’t forget the beer bottle inspection with 10 per second about 20 years ago.