• SharkieSophie

Droning On

We are living in a technological era... There are constantly new inventions, new discoveries, new products to buy, new apps to download. Every day we interact with multiple electronic devices: call phones, iPads, FitBits, smart TVs, games consoles, laptop computers... Electronics are now our social lives and our past times. Take the drone for example. When they came on the market a few years ago, we found a whole new hobby - a new way to sightsee, to film sports, to make videos. But the drone has become much, much more than just a novelty toy to scientists. Drones can have applications for tracking wildlife, preserving endangered species, and monitoring protected areas, and recent work has shown they could even save people's lives from deadly encounters with sharks!


Image of a shark identified and filmed by an unsupervised drone (Nolan et al, 2021)

Drones Have Become Vital for Scientific Research

Drones (also known as unmanned aerial vehicles or UAVs) are small aerial devices, operated remotely either by sight, or by viewing footage captured from onboard cameras via a laptop or smart phone. They are relatively inexpensive, and light weight and easily transportable (Butcher et al, 2021).

Shark being filmed by a drone (Colefax, 2020)

Since they came onto the market only a few years ago, drones have already become invaluable to scientists for capturing footage of elusive animals that would be otherwise difficult to see (Butcher et al, 2021).


For marine biologists, drones have been very handy for viewing marine animals without ever even needing to launch a boat. They can be flown over the surface of the water and any footage they capture can be analysed to see if any animals of interest, like sharks, have been spotted. Drones have already been used to collect data to study sharks all over the world (Butcher et al, 2021). To learn more you can check out Attack of the Drones.


Locations of different shark research studies utilising drones (Butcher et al, 2021)

AI is Being Developed to Recognise Sharks in Drone Footage

A surf life saver drone pilot in Australia (Butcher et al, 2021)

Drones have also been used to spot sharks in the water at popular public beaches. However, one of the drawbacks of drones is that it can be very time-consuming to trawl through all the footage and identify areas of interest - scientists must watch through endless frames of empty ocean and life guards must watch the feed at all times to be on the lookout for a shark (Nolan et al, 2021).


Therefore, computer engineers are working on creating artificial intelligence (AI), which is capable of learning to recognise particular shapes (known as "image recognition""), to then scan through footage and highlight any frames where certain animals are present (Nolan et al, 2021).

Bounding boxes a computer model added to drone footage to highlight the presence of sharks (Nolan et al, 2021)

To make this possible the computer code must be taught to recognise an animal of interest (this is known as "machine learning"). For instance, recently a team in California, used the largest dataset of its kind to train their new system! They ran their model through thousands of images that contained a shark, so that the software could learn to identify the shape of a shark from different angles, and under different lighting and ocean conditions. Their model was trained also recognise seals, tuna, surfers, paddleboarders, swimmers and boats. When they asked the system to scan through new drone footage, it was then capable of flagging frames where specific animals of interest were spotted. Not only does this allow scientists to jump straight to specific segments of useful video, but the system also circles individual animals with a "bounding box", making it easy to track them throughout the subsequent footage (Nolan et al, 2021).


Drone image of a shark near to a surfer (Nolan et al, 2021)

Drones Could Allow the Public to be involved with Scientific Research!

What is remarkable about this work, is that it was performed using consumer-grade drones and cameras, and yet was very successful at spotting sharks. The technology is still under development, but it is feasible to hope that the software could be available to the public some time in the near future. This would mean that any person in possession of a drone could could report their shark sightings and become a part of a global scientific data collection scheme (this is known as "citizen science"). Submission of videos from the public could be used to monitor populations of endangered species, to learn more about behaviour and ecology, or even to track individual animals! This continuous stream of low-cost data collection could render time-consuming, expensive scientific surveys completely redundant and revolutionise the way scientists collect data (Butcher et al, 2021).



Drones Could Stop Shark Attacks!

But this technology could also give back to the public... By helping to prevent deadly human-shark interactions!


Several different teams around the world are now working on developing AI which can scan drone footage for sharks in real-time (Gorkin et al, 2020, Butcher et al, 2021).


For example, the Sharkeye project in Australia (not to be confused with a similar project, also called Sharkeye, in the USA), surveils popular public beaches using blimps. The footage is streamed to a server, where the AI scans the video for sharks and then sends out an alert to the associated SharkMate app, to inform people that a shark is nearby. Alerts can even be synced with smart watches, so that surfers are able to receive them when they are on the water! In field trials it took only 1 minute from detection to alert the smart watch wearer to get out of the water! (Gorkin et al, 2020).

The process of shark detection and alerts by the Sharkeye project (Gorkin et al, 2020)
Drone AI can recognise sharks & surfers (Gorkin et al, 2020)

There is still some room for improvement in the system. The researchers reported a 68.7% success rate in spotting a shark, which they would like to increase. They also noted that environmental conditions had an impact on accuracy, with weather, glare and wave activity all affecting the algorithm (Gorkin et al, 2020).


However, this work shows us that it is very possible to implement a real-time warning system, that is completely autonomous (Gorkin et al, 2020).



Drone Surveillance Could Replace Destructive Beach Nets!

If drones could be used to protect ocean users from interactions with potentially dangerous sharks, this could mean it might be possible to remove beach nets! These are responsible for the deaths of thousands of sharks (and many other marine creatures!), including endangered species every year! With how far drone technology has come in such a short space of time, maybe in the very near future, we can remove destructive beach nets and replace them with drone surveillance! If so, drones could be the hero that protects both human and animals alike!


"Automated shark detection and tracking from drones could be an unremarkable and trusted presence at public beaches in the near future"

- Butcher et al, 2021




References

Butcher PA, Colefax AP, Gorkin RA III, Kajiura SM, López NA, Mourier J, Purcell CR, Skomal GB, Tucker JP, Walsh AJ, Williamson JE & Raoult V (2021). The drone revolution of shark science: A review. Drones, 5, 8. Access online.


Colefax AP (2020). Developing the Use of Drones for Non-Destructive Shark Management and Beach Safety. Southern Cross University, Doctoral Thesis. https://doi.org/10.25918/thesis.55. Access online.


Gorkin R III, Adams K. Berryman MJ, Aubin S, Li W, Davis AR, Barthelemy J (2020). Sharkeye: Real-time autonomous personal shark alerting via aerial surveillance. Drones, 4: 18. Access online.


Nolan G, Kurfess FJ, Gounder KA, Tan D, Daly C & Skae C (2021). Deep learning at a distance: Remotely working to surveil sharks. ASEE Annual Conference 2021, 34739. Access online.


By Sophie A. Maycock for SharkSpeak.

131 views0 comments

Recent Posts

See All