AI-Powered Plastic Detection and Monitoring 🌊

Using machine learning to identify plastic pollution

Did you know that approximately 9,000,000 tons of plastic enters our oceans annually? This plastic collects in large amounts and slowly breaks down, impacting more than 600 marine species while introducing toxic pollutants to their habitat, further aggravating the food chain.

The Status Quo

With a large amount of plastic being discarded every year, the majority of our plastic ends up in bodies of water, particular our oceans, impacting our ecosystem and marine life. Manually removing this plastic would be extremely costly, time-consuming and can lead to large amounts of new carbon emissions. The detection of ocean plastic pollution is largely reliant on tedious site surveys and visual identification of images. Since the late 2010s, we have used and researched AI and computer vision and make use of deep-learning algorithms to address this issue. Researchers have been developing techniques using machine learning to simplify this process. In particular, using deep-learning through CNN (convolutional neural network), which is a type of deep learning neural network architecture used in computer vision tasks for image recognition and processing. CNNs are designed to process data that has a grid-like structure, such as images and are inspired by the structure and function of the visual cortex in the human brain

The experimental setup using a GoPro connected to a power bank

Satellites, planes, balloons, drones and boats have thus been equipped with computer vision systems trained to detect plastic waste, and in theory, these innovations will pave the way for more regular detection over larger areas. However, they are still beset by a number of constraints.

In 2021, several of the responses to a European Space Agency (ESA) call for tenders for projects to detect marine plastic litter adopted this approach. In the same year, the non-governmental organisation The Ocean Cleanup announced that it had developed its own artificial intelligence surveillance and mapping tool

The Ocean Cleanup

The Ocean Cleanup uses machine learning to identify plastic pollution in rivers and simulate how it moves in the ocean. These insights power passive cleanup systems to help remove plastic that impacts our ecosystems.

Ocean Cleanup uses machine learning to identify plastic pollution in rivers and simulate how it moves in the ocean. These insights power passive cleanup systems to help remove plastic that impacts our ecosystems

They used two cleanup solutions — one for rivers and one for oceans — to remove tons of plastic from the world’s waterways. To locate plastic debris, they utilized bridge-mounted cameras to photograph floating objects and identify plastics. The Interceptor, an autonomous collection unit, is positioned to collect the plastic for removal. In the ocean, sensors attached to passive cleanup systems collect data on winds and currents. An algorithm runs simulations to show how these cleanup systems move through the ocean.

The Ocean Cleanup plans to combine these two systems at scale to reduce ocean plastic by 90% by 2040.

Training the algorithm

For AI to work, it needs to be trained to conduct a specific task. Training for AI object detection requires vast amounts of input images — the more you include, the more accurate the software becomes.

Object detection in images operates through an artificial neural network, which comprises a series of mathematical equations with distinct settings referred to as weights which helps the neural network learn to detect the object based on examples that they are trained on.

To provide the required training images, they labeled roughly 4,000 example objects in photos from our previous missions on the Aerial Expedition (2016) and their trip with System 001 (‘Wilson’) in 2018. After this, they transformed the images so that they could present them as new objects for the AI training. This process is also known as data augmentation, resulting in a dataset of 18,589 images that are ready to use.

Typical detections by the algorithm. Several verified objects after manual sorting and elimination of duplicates

Then, they located the detected object using GPS coordinate of the photograph it was detected in and these photos were grouped into different sectors. and divide that by the area of that section and used this to make a map showing where there’s a lot of plastic in the ocean.

Numerical concentrations of macroplastic (> 50 cm) are indicated by circle size and labels from the Yolov5 results. Black lines indicate photo footprints during mission NPM3. The gaps are mainly caused by transit during nighttime, at which no photos were recorded. The light- and dark-blue shaded polygons indicate the estimated outer and inner areas of the GPGP, respectively

Other Efforts

Besides the Ocean Cleanup, there are also other organizations and startups that are using AI to identify and classify plastic waste, to develop autonomous cleaning technologies, and raise awareness about the issue of ocean plastic pollution. By leveraging AI, they aim to accelerate the process of cleaning up our oceans and reducing the impact of plastic waste on marine ecosystems.

• AI is being utilized in Autonomous Underwater Vehicles (AUVs) and Unmanned Surface Vehicles (USVs) to clean up plastic and other forms of pollution. USVs are robots designed for the efficient and cost effective collection of commercial ocean data on the surface. AUVs are robots pre-programmed to collect data from specific parts of the deep ocean while scientists conduct research on board a ship or land.
• Razer and Clearbot: Razer, a tech company, partnered with marine waste cleaning startup Clearbot to advance the use of AI and robotics in reducing ocean pollution. Their goal is to develop innovative AI monitoring techniques and robotic technologies to identify and clean up plastic pollution in oceans and waterways.
• Gringgo Indonesia Foundation: This startup in Indonesia is using AI and image recognition technology to improve plastic recycling rates and reduce ocean plastic pollution. Waste workers can take a photo of trash, and the AI-powered tool will identify the items and classify them for better waste management
• Open Ocean Engineering: This Hong Kong startup has developed Clearbot Neo, an AI-enabled robotic boat that autonomously collects floating garbage in rivers and industrial waterways to prevent it from reaching the ocean. The boat uses AI to identify and collect tons of plastic waste

Our earth’s oceans are to be threatened by plastic pollution, amongst many other environmental threats. But rather than letting our oceans fill with more and more trash, we should try to consume smartly, manage our waste, and support ocean conservation groups. And as for the millions of tons of waste that already circulates through the oceans, many individuals are leveraging innovative AI technology to help clean up our mess.

Sources:

• https://hellofuture.orange.com/en/machine-learning-to-combat-ocean-plastic-pollution/
• https://www.xyonix.com/blog/using-ai-to-identify-plastic-pollution-and-expedite-ocean-cleanup
• https://www.aitimejournal.com/how-ai-helps-clean-oceans-from-plastics/#:~:text=In%20a%20nutshell%2C%20AI%20can,and%20substances%20that%20harm%20nature.
• https://www.microsoft.com/en-us/ai/ai-for-earth-the-ocean-cleanup
• https://theoceancleanup.com/
• https://singularityhub.com/2018/05/20/5-organizations-using-cool-tech-solutions-and-research-to-clean-up-the-oceans/
• https://news.microsoft.com/apac/features/this-ai-enabled-robotic-boat-cleans-up-harbors-and-rivers-to-keep-trash-out-of-the-ocean/
• https://theoceancleanup.com/updates/using-artificial-intelligence-to-monitor-plastic-density-in-the-ocean/