Mapping the plastic using UAV imagery
Article

Mapping the plastic using UAV imagery

Helping communities address the global plastic waste crisis

In response to the devastating effects of plastic on the Earth’s oceans, and specifically on waterways and coastal environments, the International Federation of Surveyors (FIG) has formed the Mapping the Plastic Working Group (WG 4.3). It has developed a plastic mapping solution, based on deep learning algorithms in conjunction with UAV-captured imagery, aimed at dealing with the problem before it reaches the ocean. The solution is already being used in Ghana, where several communities are severely impacted by the effects of plastic pollution.

Is the threat from plastic really apocalyptic? Surely it doesn’t compare to climate change or a worldwide pandemic? The answers, unfortunately, are yes, and it does (Campbell, 2022). Plastic is everywhere. It affects the daily lives of almost everyone on the planet, in all corners of the globe. Almost every piece of plastic ever made is still on our planet in one form or another, and 75% of it is now waste, with most of it discarded into landfills or dumped into marine environments (Fava, 2022). The United Nations Environment Programme estimates that only 9% of the nine billion tonnes of plastic produced globally has been recycled. As a result, each year, more than eight million tonnes of plastic end up in our oceans. This equates to around 15 tonnes of plastic entering our oceans every minute (UNEP, 2024), with 80% of all litter in our oceans now made of plastic. Without action, the World Wide Fund for Nature (WWF) estimates that, by 2050, there will be more plastic in the sea than fish by weight (WWF, 2022).

Figure 1: The UNEP 2023 report ‘Turning off the Tap’ illustrates the problem, based on a snapshot of 2020 data, in million metric tonnes (MMt).

Moreover, besides being difficult to recycle, slow to decay, and expensive and polluting to burn, plastic breaks down into microplastics (particles smaller than 4.75mm in diameter) that enter the food chain and cause harm to animals and potentially humans (McVeigh, 2022). There are an estimated 14 million tonnes of them residing on the sea floor (Barrett et al., 2020). Rivers are a recognized contributor to the ocean plastic problem. Plastic litter is predominantly concentrated on banks, coastal beaches and in the upper limits of surface water bodies during the transportation process. Researchers estimate that ten major river systems carry more than 80% of the plastic waste that ends up in the Earth’s oceans, with eight of them located in Asia (Schmidt et al., 2017; Figure 2).

The international surveying community’s response

As the international surveying and spatial sciences community’s response to this huge global problem, the International Federation of Surveyors (FIG) – representing the interests of surveyors in over 120 countries – has formed the Mapping the Plastic Working Group (WG 4.3) to help combat the global plastic waste crisis. It is a combined initiative of FIG Young Surveyors Network and Commission 4 (Hydrography). Given the GIS, remote sensing, hydrographic surveying, project management and overall measurement science skillsets within the group, the WG 4.3 focus is on the plastic waste transported in waterways and along coastlines. The aim is to find ways of dealing with the problem before it reaches the ocean.

Figure 2: 80% of the plastic polluting the world’s oceans comes from these ten rivers. (Image courtesy: Schmidt et al., 2017)

Currently, much of the plastic waste data has been obtained from large-scale empirical estimates or from detailed plastic litter surveys, which are confined to relatively small areas. The numerous global estimations of plastic debris entering oceans annually are typically based on local or regional-scale surveys, and vary from 250,000 tons (Erikson et al., 2014) to 4.8-12.7 million tons (Jambeck et al., 2015). Therefore, the amount of plastic in the world’s oceans remains poorly understood, resulting in a knowledge gap in terms of the temporal and spatial distribution of plastics, degradation and beach processes.

Several efforts have been made to establish a standardized monitoring methodology, such as Oslo and Paris Conventions (OSPAR) (OSPAR, 2020), Commonwealth Scientific and Industrial Research Organization (CSIRO) (Hardesty et al., 2016), National Oceanic and Atmospheric Administration (NOAA) (Opfer et al., 2012) and United Nations Environment Programme/Intergovernmental Oceanographic Commission (UNEP/IOC) (Cheshire et al., 2009). Those methodologies are based on traditional beach monitoring by visual counting of plastic pieces along transects. However, the accuracy of the beach survey results is dependent upon the skill of the observer, and the differences in these protocols make the integration and comparison of beach litter survey data difficult. They are also time- and labour-intensive, and only measure those items discarded during transportation rather than the amount of plastic being conveyed within a river system.

New research tackles data issues

Potentially resolving these issues, new research has shown that remote sensing data from satellites and airborne platforms can be a reliable source of information over large geographic areas. Assessment of the spatial extent and variability of plastic is possible due to the unique spectral signature of polymers in the near-infrared part of the electromagnetic spectrum. Therefore, research by WG 4.3 members at the University of Banja Luka in Bosnia and Herzegovina and the University of Novi Sad in Serbia has concentrated on distinguishing plastics from surrounding litter/debris classes using remote sensing techniques.

The initial research focused on the analysis of satellite imagery. An object-pixel-based algorithm for mapping plastic distribution in surface water using red, green, blue and multi-spectral images from high-resolution WorldView2 satellite images was developed. The paper was subsequently published and presented at the FIG Working Week held in Hanoi, Vietnam, in May 2019 (Jakovljevic et al., 2019). This research was subsequently refined to focus on higher-accuracy plastic detection over smaller geographic areas (Figure 3) using imagery captured by uncrewed aerial vehicles (UAVs or ‘drones’). To meet internationally recognized litter assessment frameworks, the aim was to successfully develop a surveying and mapping solution to accurately detect, extract and classify floating and land-based plastic particles as small as 0.01m2 (Jakovljevic et al., 2020).

Figure 3: Workflow used in this study.

Solution based on UAV imagery and deep learning

The resulting solution uses deep learning algorithms to extract floating and land-based plastic from high-resolution UAV-captured images of ‘hotspot’ locations. Customizable flight routes at low-level altitudes in combination with new algorithms, such as the structure from motion (SfM) algorithm, for photogrammetric processing provide a cost-effective solution for the acquisition of geospatial data. UAVs provide the appropriate spatial and temporal resolution to produce suitable data for mapping plastic.

For pre-processing, orthophotos are generated from the captured aerial imagery of each UAV survey. The pixel classes are delineated and labelled, and then merged using multi-resolution segmentation algorithms to create non-overlapping polygons. As this research is the first attempt to map floating plastic data from UAV imagery, previous ground-truth data is not available and the delineation, classification and labelling is done manually. It is expected that the pre-processing time will reduce as the polymer ‘libraries’ grow, and this process becomes increasingly automated.

Figure 4: Ground-truth data and results of the classification using the four tested models for detecting different plastic materials, located underwater (a) and above water (b–d).

Instead of time-consuming and labor-intensive methods largely based on visual interpretation and manual labeling of plastic pieces, the WG 4.3 solution employs the U-Net architecture (specifically, the ResUNet50 algorithm) with an encoder-decoder structure for image classification, including automatic classification, object detection, and semantic segmentation. The encoder generates feature maps with high-level semantic information but low resolution, while the decoder upsamples these feature maps to retrieve spatial details and achieve fine-scaled segmentation results (Jakovljevic et al., 2020). This end-to-end semantic segmentation model, based on the U-Net deep learning architecture, is used for classifying both floating and land-based plastic. The encoder abstracts pixel information, and the decoder extracts plastic from orthophotos.

Figure 5: The solution was tested at Lake Balkana (left) and the Crna Rijeka river (right).

Testing in Bosnia and Herzegovina

To test and refine the plastic mapping solution, UAV imagery was processed from two sites in Bosnia and Herzegovina (Figure 6). One, at the confluence of the Crna Rijeka and Drina rivers, was chosen because it is inundated with plastic and other waste. The other, at Lake Balkana, is a pristine deep-water lake. These sites were selected to test the algorithm’s ability to differentiate floating and submerged plastic from the surrounding water body, and to test its ability to differentiate floating and land-based plastic from other litter (Figure 5).

The algorithm detected plastic accurately in shallow water, which is a challenging environment for mapping plastic because the presence of the river bed increases water reflectance. Unexpectedly, the algorithm accurately extracted the plastic pieces that were omitted from the training data, showing good generalization abilities. The model also demonstrated its ability to detect plastic not just in water, but also on land.

Figure 6: The algorithm successfully extracted and classified different plastic litter types (left) from the orthophoto (right) in the Crna Rijeka river.

Plastic mapping in action in Ghana

An example of this plastic mapping solution in action is in Ghana, where plastic pollution and other litter is severely impacting communities, especially along the coastline (Figure 7). The government has committed to addressing the problem and has implemented a National Plastic Action Partnership between the government, experts, civil society, community organizations and the private sector. The Ministry of Environment, Science, Technology & Innovation (MESTI) plays a key role in its implementation.

There is infrastructure available to recycle selected plastic waste in Ghana and, currently, informal recyclers scour the beaches and other locations for items of value, such as plastic bottles. Payment is on a per-weight basis. However, the waste that has no value is left behind.

Figure 7: Korle Gonno Beach, Accra, Ghana, on 8 May 2024. (Image courtesy: Naa Dedei Tagoe)

Government initiative: ‘Plastic not Seen’

The preliminary UAV surveys undertaken at Korle Gonno Beach in Accra by students from the University of Mines and Technology, Takoradi, were assessed under the supervision of Dr Naa Dedei Tagoe and processed by WG 4.3. Because the imagery and processed results accurately depict the situation at each location at any given time, there is no basis for misunderstanding or dispute about the amount of plastic seen or not seen by any of the parties involved. The scientific rigour of the data capture and processing has gained the confidence of all participants in Ghana, fostering a concerted effort for change.

Based on this outcome, MESTI is now implementing the ‘Plastic Not Seen’ initiative (Figure 8). This is an expansion of the current informal recycling process, whereby local people are assigned a portion of coastline adjacent to their settlement and paid to clean and maintain it. The work programme for using UAV-captured imagery to classify the type and extent of the plastic and other waste at pre-determined locations is currently being finalized. Initial UAV surveys at each location will establish the baseline waste levels, with subsequent monitoring surveys undertaken to determine the level of compliance with the agreed clean-up plan.. The baseline and monitoring surveys will be undertaken locally, with the captured imagery being processed remotely – at least initially – by WG 4.3.

Figure 8: Informal coastal settlements will play a major role, and be a principal beneficiary of, the Plastic not Seen initiative. (Image courtesy: Torben Lund Christensen, Danish Association of Surveyors

Conclusion

Floating plastic and other pieces of litter are predominantly concentrated on the surface or within the upper limits of a water body. The deep learning algorithms in the WG 4.3 solution have been trained to extract plastic and other litter from the surrounding water (salt or fresh) by differentiating the spectral signatures of the plastic and the water body. This methodology can be used to map plastic waste in marine environments, including the so-called ocean gyre ‘garbage patches’, with UAVs deployed from a support vessel and the survey results processed onboard or remotely. Although initially developed to map floating plastic in rivers and waterways, this mapping solution has also proven effective in mapping land-based plastic waste and other litter, thus broadening its scope to environments such as beaches and riverbanks.

Accurate knowledge of the problem is a necessary prerequisite to finding its solution. This technology represents a significant breakthrough, enabling the policymakers of small island developing states (SIDS) to better understand the extent of the problems they face cost-effectively. The extent of each survey is limited only by the operational parameters of the UAV, and the only field personnel required are a suitably qualified pilot and spotters. As a result, this mapping solution is relatively inexpensive. The algorithms can be modified to identify other features, and the orthophotos can be processed remotely or locally depending on the available infrastructure.

The data informs decision-making, and provides affected communities – including indigenous communities – with accurate information on critical climate change indicators such as sea-level rise or deforestation. It can also inform cases for equity and economic justice in international forums, and ongoing monitoring ensures adherence to international agreements. Therefore, although it does not directly address the root causes of coastal pollution, this technology is an innovative and sophisticated step towards plastic waste eradication at a community level.

Further reading

On The Global War Against Plastics, Gordon Campbell

Jakovljevic, G., Govedarica, M. and Alvarez-Taboada, F. (2020). A Deep Learning Model for Automatic Plastic Mapping Using Unmanned Aerial Vehicle (UAV) Data. Remote Sens., 12, 1515. https://doi.org/10.3390/rs12091515

Schmidt, C., Krauth, T. and Wagner, S. (2017). Export of Plastic Debris by Rivers into the Sea. Environ Sci Technol., 51(21), 12246–12253. https://doi.org/10.1021/acs.est.7b02368

Eriksen, M.; Lebreton, L.C.M.; Carson, H.S.; Thiel, M.; Moore, C.J.; Borerro, J.C.; Galgani, F.; Ryan, P.G.; Reisser, J. Plastic pollution in the world's oceans: More than 5 trillion plastic pieces weighing over 250,000 tons afloat at sea. Plos One 2014, 9, e111913, doi:10.1371/journal.pone.0111913.

Figure 9: UAV demonstration at Korle Gonno Beach, courtesy of the Ghana Lands Commission, in May 2024. (Image courtesy: Torben Lund Christensen, Danish Association of Surveyors)
Geomatics Newsletter

Value staying current with geomatics?

Stay on the map with our expertly curated newsletters.

We provide educational insights, industry updates, and inspiring stories to help you learn, grow, and reach your full potential in your field. Don't miss out - subscribe today and ensure you're always informed, educated, and inspired.

Choose your newsletter(s)