Discarded plastic waste is a massive problem that pollutes our rivers, streams, and oceans and impacts landfills. Attempts to recycle plastic materials have been met with mixed results due to the sheer volume of plastics produced by industry, the varying types of plastics in use, consumer confusion (around what’s recyclable and what isn’t), and the complexity involved in efficiently identifying, sorting, and processing the plastic waste that enters recycling facilities.
As we explore in this Advisor, researchers, solutions providers, and recycling companies are turning to AI, computer vision, and robotics to implement new systems designed to optimize the plastic recycling process.
Optimizing Plastic Recycling
Computer vision, in conjunction with machine learning (ML) and robotic mechanical sorters, plays a crucial role in optimizing plastic recycling by enhancing the efficiency and profitability of the recycling process. This is achieved via improved sorting accuracy, increased process efficiency, optimized recovery of valuable materials, the ability to process/recycle more complex types of plastic materials, and data-driven decision-making.
Improved Sorting Accuracy
One of the biggest problems encountered in the recycling process involves sorting the different types of plastics entering the waste stream. Computer vision systems can quickly identify and sort various types of plastics, facilitating better recycling and reuse. These systems employ advanced cameras placed above conveyor belts and utilize ML algorithms trained to identify plastic materials by shape, color, and even labels as they move through the various stages of the recycling process.
Increased Process Efficiency
By automating the sorting process, computer vision systems can significantly reduce the time required for sorting plastic waste compared to manual sorting. This helps optimize throughput, offering greater operational efficiency in recycling centers.
Unlike manual sorting, which is often inconsistent and error-prone, computer vision systems provide a consistent and reliable means for identifying and separating materials. This is key; consistent and accurate identification and categorization are essential for sorting plastics because it helps ensure a higher quality of recycled product. This is beneficial for both environmental and economic sustainability because it directly impacts the usefulness and value of recycled materials.
Recovering More Materials
Computer vision systems, when combined with ML techniques and AI analytics, can distinguish between different types of plastics, including compostable and biodegradable plastics. Such applications utilize highly sensitive imaging techniques and advanced classification models designed to facilitate accurate and automated sorting of various types of plastic.
This is a particularly important development: although the use of compostable plastics in society is increasing, their drawback is that they can mix with and contaminate other types of (conventional) plastic waste during recycling if not properly separated from the waste stream.
Handling Complex Materials
Vision systems have evolved considerably. They can now sort plastics that were previously difficult to identify using traditional methods; for example, black plastics, which are commonly used in automotive, food packaging, electronics, and bagging. With the incorporation of hyperspectoral or multispectoral imaging capabilities, these systems can now sort black plastics, which helps improve the quality of plastic recycling by ensuring that such materials don’t end up discarded in landfills.
We are also seeing the development of vision-based systems that can identify so-called “soft plastics,” like cling wrap and plastic bags, in the waste stream. This is especially important because soft plastics are difficult to recycle using automated recycling methods due to their tendency to get tangled in waste sorting and separation machines, causing mechanical breakdowns and contaminating other recyclables.
Data-Driven Decisions
Data generated by computer vision systems can be analyzed to provide recyclers with insights into the number and mass of materials processed and to identify trends they can use to optimize processing and (ideally) increase profits. Such information is also valuable for assisting brands with the design of recyclable products.
Developing Computer Vision Systems for Plastic Recycling
Developing computer vision systems for plastic recycling requires building and training neural networks that can differentiate between the different types of plastics in use today regardless of the condition of the materials when entering a recycling center. This presents a number of challenges.
The sheer variety of plastics (i.e., shapes and compositions) can make it difficult for computer vision systems to accurately identify and classify plastic waste. For example, different plastic bottles traversing a conveyor belt may appear visually similar but actually have very different chemical compositions that require different processing for recycling. In addition, plastic items are often contaminated with food and other debris. They can also be smashed, torn, or deformed.
These variables complicate the training process because the neural net must learn to identify the different types of plastics despite such inconsistencies for the vision system to accurately perform its sorting functions. In short, poor sorting accuracy can significantly reduce the quality/value of the recycled output.
Due to these considerations, training neural nets for computer vision systems that can accurately and efficiently support recycling operations demands large, diverse training data sets representative of the varying types of recyclable plastics in use today. Moreover, neural net models used in vision-based recycling systems will require updating to keep pace with the frequently changing nature of new plastic materials entering the market.
Conclusion
Computer vision, in concert with ML and robotics, is playing a crucial role in modern plastic recycling operations by optimizing the efficiency and profitability of the recycling process. By improving the sorting process, computer vision systems can enhance the recovery rates and quality/value of recyclable materials. This prevents plastic waste from ending up in landfills and polluting the environment and makes it easier to reprocess and reuse materials, in effect contributing to a more sustainable and circular economy. Implementing computer vision systems for plastic recycling presents several challenges. However, the potential benefits of improved recycling processes are driving considerable research and innovation around the development of computer vision systems and ML techniques for optimizing plastics recycling.
In Part II of this Advisor series, we’ll examine some applications and solutions employing computer vision, AI, and robotics for plastic recycling. In the meantime, I’d like to get your opinion on the use of these technologies for plastic recycling and recycling in general. As always, your comments will be held in strict confidence. You can email me at experts@cutter.com or call +1 510 356 7299 with your comments.
