Wednesday, October 27, 2021

Smart Safe Keeping: Blending Artificial Intelligence with Sea Turtle Conservation

Author Stephen Taglieri, KwF Conservation Research Intern

More and more, drones are becoming a normal part of our future. “When we first
introduced the integration of computer science with aerospace engineering to create self aware drones it seemed like an alien concept, but over the last couple of years A.I. has advanced exponentially while drone development has expanded to many conservation studies,” says Princess Aliyah Pandolfi, Executive Director of Kashmir World Foundation (KwF).

These flying robots aid workers with daily tasks, and innovation keeps pushing technology in a direction to further help. However, drones don’t always have to help people, drones can also be used to safeguard wildlife. Conservationists are chronically underfunded and understaffed, so the use of drones can give much-needed assistance. This is especially true with sea turtle conservation.

Kashmir World Foundation (KwF) was able to introduce Pronatura, Mexico’s largest
environmental conservation group, to the perks of using drones to protect turtles.

Pandolfi created the Fly for Conservation workshop to educate researchers and biologists on the value of custom drones embedded with A.I. on the Edge. In 2018, Marista University of Mérida hosted the Fly for Conservation workshop with KwF staff teaching environmentalists on the Yucatan Peninsula how to build, program, and operate drones to survey sea turtles efficiently.

With aerial visuals, the team released 32 sea turtle hatchlings at a beach in Celestun, Mexico and ended the successful evening with a group picture to document their victory. Sea turtle conservation is usually time-consuming and energy-intensive; drones, and the innovative ways to use them, take pressure off experts so they can spend more time-saving species.

Dr. Melania López-Castro is a conservationist at Pronatura working with sea turtle populations along the Yucatan coastlines. Drones have the potential to help their efforts significantly.

Although the organization is large, Pronatura still struggles with finding enough manpower to survey all the sea turtles that nest on beaches under their jurisdiction. There are many threats to turtles in that area, and López-Castro is working to increase her organization’s ability to protect them. They are currently fighting against the effects of plastic pollution, beach erosion, and destructive human activities on the beach.

The Pronatura team walks the beach every morning and evening to find any new turtle nests. Throughout the night, researchers and volunteers work to record the success of hatchlings leaving the nest and avoiding predators on their way to the ocean. Animals such as birds, crabs, coyotes, raccoons, and sharks are all predators of sea turtle hatchlings. If the team runs into an adult turtle while looking for tracks on the beach, they will measure the animal and record whether it has been tagged before. The process is very demanding of the researchers, and it can stress the turtle as well.

This is where KwF steps in to create a more effective and minimally intrusive conservation environment for both humans and turtles. Malini Shivaram, Sean Sewell, and Langston Kosoff are all interns working with the artificial intelligence A.I. program at KwF.

They are developing a new artificial intelligent drone that can help sea turtle conservationists.

Shivaram is an Artificial Intelligence Major at Carnegie Mellon, Sewell an Aeronautical and Astronautical Engineering Major at Stanford, and Kosoff is a Computer Science Major at the University of Chicago. The sea turtle conservation drone embedded with an AI, is called MiSHELL. It is made specifically for detecting sea turtle tracks and identifying the specific species and providing the GPS location of the nest to the biologists.

KwF has partnered with Pronatura to collect data for their research and test MiSHELL’s ability to work outside of lab settings. The Convolutional Neural Network (CNN) is trained to process data in real time on a small computer onboard the drone. This type of program is best for image recognition and processing. Once the computer is connected to the flight controller and GPS on the drone, it will be used to fly over beach areas where turtles may have been. Using downward-facing cameras, MiSHELL will be able to process the video data in real time and identify sea turtle tracks as it flies over them.

Conservation teams in the area will then know where the tracks are located and can navigate to them via GPS coordinates. The project will save time finding the animals and give the Pronatura more time to record data and protect the sea turtles.

In an interview, Shivaram said that “using AI to identify wildlife in captured images or
videos could be an efficient and fairly accurate way to process data.” With MiSHELL, real-time data is sent to on-the-ground conservation teams while the drone captures imagery. The data can be sent to these teams at any time, even if they’re miles down the beach from the drone’s location. The team will know specifically where to find the tracks, so they won’t have to walk meticulously down the beach looking for turtles. MiSHELL can even identify which species of turtle created the tracks on the beach. This is done by analyzing the pattern of these tracks; all seven species of turtle leave different patterned tracks due to the size and shape of their flippers.

Creating a working, reliable AI model takes a great deal of data. KwF’s partnership with Pronatura has brought the AI-building process a long way. Researchers in Mexico have recorded videos from the drones flying over turtle tracks and sent them to KwF. These pre-recorded videos are used to train the CNN. The data is fed into the AI so it can learn how to successfully identify the tracks. Shivaram and the rest of the team all work to point out the tracks to the AI initially, and then challenge the program to find the tracks by itself. This process is called “training a convolutional neural network,” and it allows the AI to identify these patterns in the sand on its own.

This careful process takes a long time. It gives the team time to work out any other bugs, though, before fully putting their trust into the drone. For example, López-Castro observed that the sound of drones spook turtles in the area. While the AI continues to be developed, the drone team at KwF is using research from another project, Eagle Ray, to fix this issue. Eagle Ray is a different kind of drone program which employs silent thrusters, and that technology could cross over to benefit the sea turtle project (Read more about Eagle Ray).

As drone technology gets more advanced, López-Castro and her team can’t help but look into the future.

The applications of drone technology in the conservation field are seemingly endless. The Pronatura team believes that the technology could be used for more applications as they get more involved with the drones. One potential use would be to track sea turtle surface movement in clear waters off the coast. Tracking turtles using aerial drones would allow researchers to get close to turtles without noisy boats. This method could also give more information about where sea turtle breeding and feeding grounds are. Drones made for underwater environments could replace metal or satellite tagging programs while being more cost-effective.

This is a huge step in the right direction. When speaking about current tracking methods, López-Castro explains that the most effective way of tagging a turtle in the wild is to clip a metal marker to the animal’s flipper. Each marker has a unique number that identifies that specific turtle, and researchers can record where and when they see that animal again. The downside is this method does not give any data about the animal between its interactions with researchers, which means that not much is known about turtles when they’re off the beach.

Satellite tags can be used, but they are very expensive and have a high chance of falling off the turtle before enough data is taken. Experts still don’t know where sea turtle feeding grounds are, where mating sites are, or the impacts which the disappearance of sea turtles is having on the environment.

Working with what they are given, conservation organizations have been able to make the best of their resources. KwF, and partners, work to give environmentalists the technology and training to grow eco-friendly efforts around the world. The use of artificial intelligence, drones, and other state-of-the-art innovations can change the game when it comes to saving species. Having helped create a MiSHELL, Shivaram affirms that “like in many other fields, technology (especially the use of AI) will become prominent in conservation.”

Visit our website at www.kashmirworldfoundation.org

Rob Thompson
Rob Thompson is the co-founder of Falcon Foundation, a 3rd generation commercial multi engine pilot, Part 107 holder who also holds a Master of Science from James Madison University for his work in aviation system designs and technical & scientific writing. Falcon Foundation provides leading advocacy efforts in the unmanned aircraft systems industry, managing government relations, committees of association, executing legislative and regulatory strategies and creating law through the corresponding legislative committees. By working independently on advocacy issues, educating the clients on public policy issues quickly, and by engaging team members to facilitate successful results. Client policy issues will include aviation regulation, unmanned aircraft systems, Part 107 waivers, the regulatory process, and industry safety concerns. Client groups include aviation professionals, unmanned aircraft systems, and operators, both commercial and hobbyists, and non-aviation business sectors, including small business service and manufacturing sectors.