Malaysian researchers are testing drones to detect trash and notify local authorities for clean-up. — Photos: ART CHEN/The Star
Imagine having a bird’s-eye view of the ground, spotting areas where trash accumulates, often unnoticed and ignored by many. While out of sight may mean out of mind for most, some are increasingly concerned about the long-term environmental impact of uncollected waste.
“I just feel there’s a lot of waste everywhere in the city, even in areas like the beach and by the river,” says Raja Mazhar Mohar Raja Tun Mohar of the Malaysian Humanitarian Foundation (MHF), a non-governmental organisation aiming for 5% of Malaysians to become skilled, long-term volunteers by 2030.
In an interview in Subang Jaya, Selangor, Raja Mazhar explains that while the NGO has traditionally concentrated on hospice care and social support, he is becoming increasingly aware of the broader impact that environmental issues, such as climate change, can have on humanity.
“You can see changes like unpredictable weather patterns that have affected people. One of the contributing factors in climate change, I feel, is unmanaged waste. So it got me thinking – how can I play a better role here in coming up with a solution to help the authorities tackle this issue?” he says.
The United Nations in 2022 described unmanaged waste as “a hidden cause of climate change”. When trash such as organic items rots in landfills, it releases methane, a type of greenhouse gas that traps heat and ramps up global warming.
After researching and discovering the work of Dr Joanne Lim Mun Yee at the Internet of Things (IoT) lab in the School of Engineering at Monash University Malaysia, Raja Mazhar arranged a meeting with her.
Together, they developed the Drone Waste Alert Management System – 5R. Lim explains that their vision is to use drones to capture data on accumulated ground waste and send alerts to the city council for collection.
“Right now, they have to manually search for trash to collect, which could take up a lot of time. We believe this innovation can benefit city councils to perform waste collection more efficiently,” Lim adds.
Training AI to identify trash
Monash University Malaysia School of Engineering PhD student Liew Chan Yue was brought onto the project by Lim.
First, he needed to obtain a drone handling licence from the Civil Aviation Authority of Malaysia, a process Lim compares to learning and passing a car driving test. With the licence in hand, Liew began developing an AI model to identify trash from the air, starting with the collection of household waste.
“We tried to identify the most common types of waste that you will typically see in the streets and that boils down to bottles, plastic, papers and cans.
“From there, we chose a specific drone testing facility at Area 57 in Bukit Jalil, Kuala Lumpur. Flying the drone to collect datasets took about two months,” he adds.
Next, Liew moved on to labelling the datasets.
“Machine learning models need labelled datasets. So from the videos that we record, we split them into pictures and manually annotate each image. For example, if there’s plastic on the ground, we’ll draw a box around it to indicate that it’s plastic. This information is then fed into the machine learning model,” he says.
Liew explains that for the AI model to recognise items like bottles, it analyses the colour and shape of the objects. There was also significant fine-tuning required throughout the process, as they had to make adjustments to improve the model’s detection speed.
“Basically, through more programmatic methods, we were able to reduce the time it takes to detect an object by half. In terms of accuracy, we solved the issue by increasing the amount of datasets as well as using different parameters,” he says.
Overall, he says the model was trained on around 2,000 images where each image has about 20 labels.
Raja Mazhar says the painstaking process of annotating the datasets and constant fine-tuning improved the accuracy of the model, to which Liew agrees, claiming that its accuracy at identifying objects on the ground is about 90%.
A 3D-printed enclosure was used to house a computation device that runs the AI algorithm that can be connected to the Internet.
The housing, which includes a GPS and a camera, was then fitted to the drone.
A critical consideration, according to Liew, was ensuring the module wouldn’t weigh down the drone.
A mobile app was developed to monitor the AI module, allowing operators to track the drone’s movements and access real-time data on trash it detects, which is transmitted to the cloud.
“What we absolutely need on this app is a heat map that will allow the operator to see where there is a high concentration of rubbish detected,” Liew says.
Taking it to new heights
Last April, the researchers were invited to showcase their drone system to the Penang Island City Council in conjunction with the Earth Cleanup Day Event.
“We showed them the possibilities that the system can bring to help improve waste management,” Lim says, adding that the council expressed interest to know more about it.
“Essentially, the Penang Island City Council was open to the idea of the system, including the possibility of training the existing workforce to become drone operators,” she adds.
Apart from monitoring waste in hard-to-reach areas, Lim says there are plans to include a feature that detects and sends alerts about litterbugs.
At the moment, the researchers are gathering the data for this feature.
“It will be able to detect faces of people throwing rubbish, and we hope that this can be used to improve the compliance level of those utilising public areas like parks,” she adds.
Raja Mazhar recommends that offenders get a warning and be reminded that they are being monitored for compliance, proposing that public parks have signs indicating they are under surveillance.
He says drones intended for official monitoring should be painted in a distinct colour to signify their purpose.
“For example, we can inform people that green drones are meant for waste monitoring and we believe that more people will come to accept the roles that drones can play in the city,” he says.
The system can work without drones, says Lim, as the AI can also monitor CCTV cameras in the city.
According to Liew, the beauty of the project lies in its adaptability – additional hardware like infrared cameras can be integrated to enhance nighttime detection.
He says various AI models can be loaded to accomplish different objectives.
“For example, we’re looking to use this system to detect and send alerts about potholes as well,” he says, adding that they are also going to add 5G connectivity.
According to Lim, the next phase of the project involves testing the system with a larger drone that can operate for a longer duration. Meanwhile, Raja Mazhar plans to approach more city councils to showcase the system, with the hope that they will consider the project for implementation.
They also plan to pilot the project in the city, but Lim says this would require additional approval from the Civil Aviation Authority of Malaysia.
“In Malaysia, we still have very strict regulations on where and when drones can be deployed. We understand the need for regulations, so we will work harder towards seeking their support for the project,” she adds.
Lim believes drones will play a bigger role in easing many processes in the future, including food delivery.
In the United States, places like New York City have deployed patrol drones to monitor sharks and struggling swimmers.
Recently, it was reported that the Singapore Police Force deployed multiple drones with speakers and blinkers as part of a safety enhancement feature to monitor the crowd at a New Year countdown event.
Lim says it’s only a matter of time before drones are deployed for various purposes in the country.
Upon welcoming his first grandchild this year, Raja Mazhar began reflecting on the future. This fresh perspective, combined with his collaboration with Lim, was what served as a catalyst for developing a drone-based waste alert management system.
“I’ve been looking at the effects of climate change, and I’m worried: What if there are not enough food resources for the next generation? I hope to do what I can now so they don’t suffer from the mistakes of my generation,” he says.