Search other topycs

Categories

Satellite images + AI: The counterattack to the plastic invasion?

Satellite images + AI: The counterattack to the plastic invasion?

Satellite images The counterattack to the plastic invasion

According to the UN (2017), every year 8 million tons of plastic end up in the oceans. To measure this, it is equivalent to annually discarding the mass of more than 40,000 Antarctic blue whales (the largest animal in the world), taking into account that the estimated population of this subspecies in 2018 was 3,000. So yes, the problem is colossal.

According to Dr. Biermann et al. (2020), certain characteristics such as the durability and resistance of this material (which are considered advantages in different industries), now play against us, because these wastes end up stagnating in high impact areas or affecting the beings that inhabit them. the coasts and the sea, since many wastes are ingested by these or end up strangling them, in addition to seriously affecting all ecosystems.

The detection of macroplastics (plastic fragments with a diameter greater than 5 mm) is a priority, considering that if these cannot be located and removed before they fragment, the subsequent extraction of this waste becomes impossible. On the other hand, Dr. Beaumont et al. (2019), who has more than 17 years of experience studying the intersection between socioeconomics and natural sciences, suggests that there is a considerable socioeconomic impact given that it affects agriculture and tourism.

How to counteract contamination with satellite images?

The potential of satellite images to solve multiple problems has always been known thanks to the perspective that they can provide; However, prior to the launch of the Sentinel-2A and 2-B satellites in 2015, there was no opportunity to develop effective solutions for locating plastics as the satellites, until then, had one of the following problems:

  • The resolution was not sufficient for the detection of elements of reduced dimensions, this means that each pixel represented a very wide area.
  • The number of spectral bands provided was not ideal to be able to obtain adequate data, since it is related to the amount of information that is acquired from a scene with respect to the electromagnetic spectrum. As can be seen in Image 1, a band refers to a range of spectrum values (X-axis) which is of great importance when you want to locate a material or phenomenon.

These obstacles were overcome with the Sentinel 2, which have a resolution of 10 m and 12 spectral bands.

Illustration of spectral bands
Img 1Illustration of spectral bands by Satellite Imaging Corporation, 2014.

Some proposals for plastics detection

There are projects with significant contributions such as the Plastic Litter Project, of the Marine Remote Sensing Group of the University of the Aegean, which is located in Mytilene, on the island of Lesbos. In 2018, this project released floating debris composed of plastic bags, bottles or fishing nets in a controlled manner, as seen in Image 2, which served to work as an ideal environment for the identification of floating debris by satellite images.

Image captured by drones
Img. 2 Image captured by drones on June 7, 2018 at Tsamakia beach, Greece.

In addition, this group carried out a characterization of the floating plastic (Topouzelis et al., 2019), obtaining an approximation of its spectral signature. This has to do with the idea that the radiation that a material reflects in the electromagnetic spectrum is unique. For example, Image 3 shows the spectral signature for water, land and vegetation, with reflectance on the Y axis and the electromagnetic spectrum values on the X axis, where our eyes can only receive information from the former 3 bands (blue, green and red), but we are missing a lot more information.

Spectral signature for water, land and vegetation
Img. 3 Spectral signature for water, land and vegetation by SEOS, s.f.

Using the characterization performed, Dr. BIermann suggests an index for the detection of floating debris called the FDI (Floating Debris Index). To test this index we obtain the unfiltered visualization of Tsamakia beach (Image 4) with the Sentinel HUB platform (in which custom filters can be applied using Javascript as indicated in the guide) and in Image 5 we apply the filter obtained with the proposed FDI index. Although it is evident that plastics are highlighted, it must be taken into account that this index is taking low values, that is, it is not conclusive and when tested with chaotic scenarios, it does not discriminate other materials.

Satellite image of validation plastic
Img. 4 Satellite image of validation plastic. Greece from June 07, 2018.
Satellite image with FDI filter applied
Img. 5 Satellite image with FDI filter applied.

On the other hand, the Sentinel HUB platform has a large number of predetermined filters, in addition, it runs contests to encourage the creation of these, which are included in a if they turn out to be useful. One of these is the Ocean Plastic Detector, which seeks to detect plastics in the ocean, although, as seen in Image 6, (in the same validation case above) it seems to work correctly. Image 7 shows that under certain conditions, both in the FDI filter and in the Ocean Plastic Detector, cases of false positives are observed, that is, they incorrectly highlight the possible plastics.

Satellite image with Ocean Plastic Detector filter applied
Img. 6 Satellite image with Ocean Plastic Detector filter applied.
False positives from FDI
Img. 7 False positives from FDI filters.
False positives from 'Ocean Plastic Detector'
Img. 7 False positives from 'Ocean Plastic Detector'.

Endless battle

Satellite images with ia Endless battle

Clearly there are very interesting proposals for the detection of plastics, however, there is still a long way to go, given that the proposed filters have a lot of false alarms, and that is where artificial intelligence has an opportunity to go beyond human intuition and usual science.

The threat is latent from plastic, according to the UN (2017) it is estimated that by the year 2050 the oceans will contain more plastic than fish, in addition, that 99% of seabirds will have ingested it. Therefore, even if the best plastics detection model is achieved, our responsibility as inhabitants of this planet is to attack the pollution problem by generating environmental awareness, without waiting for a tool to solve our pending accounts, because if so, it is an endless battle.

References

  • Beaumont, N. J., Aanesen, M., Austen, M. C., Börger, T., Clark, J. R., Cole, M., . . . Wyles, K. J. (2019). Global ecological, social and economic impacts of marine plastic. Marine Pollution Bulletin, 142, 189-195. doi:10.1016/j.marpolbul.2019.03.022
  • Biermann, L., Clewley, D., Martinez-Vicente, V., & Topouzelis, K. (2020). Finding Plastic Patches in Coastal Waters using Optical Satellite Data. Scientific Reports, 10(1). doi:10.1038/s41598-020-62298-z
  • ONU. (2017).  UN Declares War on Ocean Plastic.
  •  Satellite Imaging Corporation. (2014). WorldView-3 Satellite Sensor.
  • SEOS (s.f.). Classification Algorithms and Methods.
  • Topouzelis, K., Papakonstantinou, A., & Garaba, S. P. (2019). Detection of floating plastics from satellite and unmanned aerial systems (Plastic Litter Project 2018). International Journal of Applied Earth Observation and Geoinformation, 79, 175-183. doi:10.1016/j.jag.2019.03.011
  • WWF. (s.f.). Antarctic blue whales recovering. Recuperado de