What is computer vision and how can it help the waste industry?

Written by: Alisa Pritchard | Published:
Positive sorting- computer vision can identify and sort different types of PET plastic

When it comes to computer vision, most people think of the technology that powers self-driving cars

Alongside lidar sensors, computer vision enables the driverless car to ‘see’ and understand objects on the road to instantly speed up, turn or stop. Computer vision is an area in artificial intelligence (AI) that can analyse images and videos to identify and categorise objects.

Examples of where this technology can be used are endless, from healthcare to retail to manufacturing, resulting in the exponential growth of an already £8.7 billion market for computer vision.

When applied to waste management, recent advancements in this technology now allow machines to ‘see’ and classify waste at human-level recognition or better.

Why does this matter, and why now?

While AI technology has caught up with and overtaken human-level vision capabilities, the waste industry is also being driven forward by a number of external factors.

On the demand side, producers and manufacturers are under pressure to use recycled materials. Meanwhile, the rejection of poorly sorted waste by China and other emerging markets has increased the onus on developed countries to output high quality and high purity waste for recycling.

Against this backdrop, AI, and computer vision in particular, has already started to enable new and innovative ways to audit, recover and sort waste - from smart bins to automation in recycling plants.

How is this different from machine vision?

An easy way to understand where deep learning-based computer vision can typically be deployed is to ask: where is human vision needed to make a split-second decision?

Computer vision is a scientific domain, while machine vision is an engineering one used in industrial settings. Machine vision deals with industrial environments where light and motion are controlled. The objects are already known and the observed events are predictable. Computer vision deals with unknown objects which are often in uncontrolled environments.

Recent advances in AI have given rise to a new way to solve such complex visual tasks, in particular using deep learning (DL) which is a subfield of machine learning (ML). While both DL and ML fall under the broad category of artificial intelligence, deep learning is what powers the most human-like artificial intelligence. It is the use of models called neural networks, which are composed of many processing layers and can be trained to learn how to make intelligent decisions from known examples.

This technique inspired by neural connection in our brains, allows analysis, automatic extraction, and understanding of meaningful data from a single image to a sequence of images. The result is a computer vision solution that can exceed human-level performance.

Negative sorting- identifying impurities in the PET plastic line to remove them

Waste management is undoubtedly a complex environment, with an infinite number of ways in which waste can be manipulated. Current machine vision solutions are excellent at performing specific and predictable tasks. However, they struggle with ambiguity and variety of objects and tasks that require human (or human-level) judgement - hence the extensive manual sorting currently deployed in the industry.

As Mikela Druckman, Founder & CEO of Greyparrot, explains: “AI is changing many industries and has the potential to significantly impact waste management as well. The combination of fast advancements in computer vision and decreasing cost of robotics will drive further automation and transparency in the industry.”

At Greyparrot, we are developing deep learning-based waste recognition systems for next-generation robotics and smart systems for waste management. Current projects include the automation of quality control in MRFs and waste composition analysis at collection points.

Computer vision applications to improve resource capture

Rapid advancements in AI technology has unlocked a real opportunity for the industry to make a step-change in waste recovery efficiency and quality.

For developing nations, a lack of legacy infrastructure will see the immediate introduction of such technologies. This will allow for more advanced decentralised systems to deal with increasing waste flows.

For developed countries, the opportunity to track waste and introduce automation is clear. Moreover, pressure on consumer goods multi-nationals in the form of EPR (Extended Producer Responsibility) gives rise to an urgent need for deeper insights and transparency on waste flows along the supply chain.

Innovative players are beginning to try and apply this technology to reap the benefits of a more automated and efficient system. Ultimately this [AI and robotic application] is where recycling and waste management is going.

Alisa Pritchard is head of marketing and operations at deep learning company Greyparrot.

Greyparrot will feature at the Smart waste, smart resources conference on 16 October in Birmingham. Click here to book your ticket now.

This material is protected by MA Business Ltd copyright.
See Terms and Conditions.


Please view our Terms and Conditions before leaving a comment.