Drones, smart cameras and machine learning for remote environmental monitoring
Researchers and Unmanned Aerial Vehicle (UAV) engineers are pioneering the use of new miniaturised hyperspectral cameras to monitor the health of Australian landscapes in more detail than ever before. They can now create super fine, fingerprint-detail maps of landscapes using UAVs equipped with highly sophisticated, light- weight camera technology for remote sensing.
Working at the cutting edge of this technology, Associate Professor Felipe Gonzalez, from Queensland University of Technology (QUT) and his team have obtained aerial hyperspectral imagery of a coral reef in extraordinarily high resolution. They are gathering data about pristine Ningaloo Reef at the Ningaloo Coast World Heritage site in Western Australia, to inform future planning. At 290 kilometers, Ningaloo Reef is one of the longest and most structurally complex nearshore reefs in the world.
Hyperspectral cameras have reduced in size and are now light enough to be integrated on UAVs. Instead of capturing images in just 3 bands of the visible spectrum like normal cameras, hyperspectral cameras capture 270 bands in the visible and near-infrared portions of the spectrum, providing far more detail than the human eye can see.
“Improved detail means more accurate data and that leads to better informed decision making for land planning and conservation,” said Professor Gonzalez.
Professor Gonzalez and his team from QUT’s Institute for Future Environments (IFE) can capture very high resolution images because they are flying these cameras on small UAVs at 30–100m over the water.
The new lightweight hyperspectral cameras open many possibilities for reef monitoring and collaborations with marine researchers. IFE researchers are already working on integrating a hyperspectral camera unit into an underwater housing for marine robots.
UAVs are a cost-effective sensor platform and a complementary tool to existing satellite, manned aircraft and underwater surveys.
Large-scale, high-altitude surveys of the Ningaloo and Great Barrier reefs lack the resolution necessary to identify individual corals and this technology provides for low-altitude but highly detailed, UAV surveys.
“There has long been a trade off between scale and detail as well as cost. Because of the advance in camera technology and drone technology, we can now bridge that divide and get greater detail over vast distances at low cost,” said Professor Gonzalez.
The data gathering and analysis system is backed by a state-of the-art, IFE’s Research Engineering Facility (REF)-designed and manufactured gimbal that enables stabilised image capture from the hyperspectral camera on-board QUT’s Unmanned Aerial Vehicles.
Environmental data analysis
Using data collected by engineers in the IFE’s Research Engineering Facility (REF), Professor Gonzalez and his team are developing revolutionary software to quickly analyse the airborne hyperspectral information from the reef. This software is adaptable for a wide range of other environmental purposes, including detecting invasive plants in Western Australia and diseases in wheat crops.
Dr Dmitry Bratanov, a research engineer, and part of the team at QUT, utilised the hyperspectral UAV system to survey approximately 40 hectares of Ningaloo Reef in 30 minutes at a flight height of 100m.
This highly efficient sensing method provides the research team with a spatial resolution of approximately 15cm per pixel — more than enough detail to detect and monitor individual coral species.
As the hyperspectral camera takes images across 270 slices of the spectrum, gathering a significant amount of information, this allows for the classification of coral species, sand and algae based on their unique spectral signatures. These signatures act very much like fingerprints.
Gonzalez and his team are collaborating with marine researchers to build artificial intelligence algorithms that can automatically recognise and classify these unique signatures — the hyperspectral equivalent of a fingerprint database.
This database, identifying individual species with their unique colour signatures, will be highly valuable to all environmental researchers in the future. It will standardise the species and reduce the cost and time required to gather and analyse data.
NASA’s Coral Reef Airborne Laboratory (CORAL) mission is currently conducting hyperspectral mapping of the Great Barrier Reef at high altitude with manned aircraft. The CORAL system provides a resolution of 3.5-7.5m per pixel. QUT’s UAV system, by comparison measures at 15cm per pixel. QUT also uses a manned fixed wing aeroplane fitted with a hyperspectral camera that can capture data at 35cm per pixel from 230 metres off the ground.
The advantages to using smaller hyperspectral cameras and a UAV: it’s cost-effective, quick to deploy and flexible. A mission can be scheduled any time in strategic locations where a higher level of detail is required.
In future, comparison studies of reefs, for example between a remote pristine reef in Western Australia and a reef system under pressure from human activities along the Queensland coast, could be very useful.
For more information about the work of Professor Gonzalez and his team, go to…
RELATED STORIES
The robot eyes have it: cutting-edge tool for koala conservation
Crown-of-thorns starfish — nailed it!
Aerial robots, artificial intelligence and statistics revolutionise wildlife tracking and research
Find out more about research at QUT.
