How AI is Helping Catch Animal Poachers

Introduction

The illegal wildlife trade is a significant global issue, particularly in regions like Africa, Cambodia, and South America, where poaching threatens many endangered species. To combat this, conservationists are increasingly turning to artificial intelligence (AI) as a powerful tool. From fixed cameras hidden deep in the wilderness to drones patrolling the skies, AI is transforming how we safeguard wildlife. In addition, AI is being used to detect wildlife snares, a problem across many reserves.

Fixed Cameras: Silent Sentries in the Wilderness

One of the most efficient AI applications in anti-poaching efforts involves fixed cameras equipped with deep learning technology. For instance, TrailGuard AI, developed by the US-based environmental organisation RESOLVE and manufactured by Nightjar, uses tiny, camouflaged cameras to monitor endangered species and detect poachers. TrailGuard’s primary aim is to stop poachers before they strike, acting as a burglar alarm for national parks. Since no existing camera met their requirements, they designed one specifically for these challenging environments.

What sets TrailGuard apart is that it’s the first tiny AI-embedded camera system designed for this purpose. It is exceptionally low-power and boasts a long battery life, capable of functioning for up to 18 months in the wild. The cameras use Intel technology to transmit images to rangers in as little as 30 seconds, utilising either cellular networks or long-range radio signals. This immediate alert system allows rangers to respond swiftly to potential threats.

In the Serengeti, where TrailGuard was installed, there are only 155 rangers tasked with protecting a vast area. If the park had hundreds or even thousands of TrailGuard units, these rangers could become as effective as having 1,500 or even 15,000 rangers on the ground. The system’s ability to cover large areas with minimal manpower offers a significant boost in the fight against poaching.

TrailGuard’s custom-built AI models can be trained to detect not only human intruders but also any species of interest, from tigers to hyenas, making it a versatile tool for protecting a wide range of wildlife in different environments.

During initial tests in East Africa, TrailGuard’s AI system resulted in the arrest of over 30 poachers, and its large-scale rollout is now being trialled in tiger-rich regions of central India.

TrailGuard AI was included on TIME's List of Best Inventions changing how we live 2023.

AI Identification

There are many groups working on animal identification. Conservation AI is a research project developed at Liverpool John Moores University which uses artificial intelligence to process the large amounts of data generated by conservation tools such as camera traps and acoustic sensors placed in the wild. Conservationists have struggled to keep up with the sheer volume of data from these sources, and AI offers a solution by allowing this data to be analysed quickly and efficiently.

The project has been in development for six years and has been actively used over the past 18 months. So far, it has processed around 20 million images and has ongoing projects in various regions, including work with the Snow Leopard Trust in Asia and monitoring pangolins. Conservation AI was the first organisation to detect a pangolin in real time using AI technology. A camera in Uganda detected a pangolin, and within 20 seconds the image was processed in the UK and an alert was sent to rangers. As a result, the poacher attempting to catch the animal was intercepted.

The system works with real-time cameras connected via 3G or 4G networks, and these cameras are distributed globally to monitor wildlife and help with human-wildlife conflicts, as well as anti-poaching initiatives. The speed of AI processing gives conservationists an edge in biodiversity monitoring, allowing them to react quickly to threats. Conservation AI estimate that rangers can detect poachers 17 times faster when using AI combined with a drone compared with using a manual method.

PAWS: AI Tracks Poachers Before They Strike

The PAWS (Protection Assistant for Wildlife Security) system is being used to predict poaching hotspots. This AI technology analyses historical patrol data and environmental factors to forecast areas most likely to see poaching activity. Rangers use this predictive information to focus their patrols more effectively.

PAWS uses an area of game theory known as security games to model scenarios where rangers (defenders) must allocate limited resources to protect against poaching attacks. Security game theory is often applied in AI technologies to create strategies for optimal resource distribution. In PAWS, it allows the system to predict high-risk poaching areas by analysing historical data from the SMART (Spatial Monitoring and Reporting Tool) system, developed by the World Wildlife Foundation. 

In 2014, Milind Tambe and his team first trialled PAWS at Queen Elizabeth National Park (QENP) in Uganda, where thousands of snares are set by poachers across a 2,000 square kilometre area patrolled by just 100 rangers. PAWS divides the park into 1km squares, assigning each a risk factor based on SMART data collected over a decade. The system suggests patrol routes through the highest-risk areas, factoring in variables such as time of year, trails, rivers, and weather. 

During a six-month test, rangers followed both high- and low-risk routes. “Where PAWS predicted higher risk, more snares were found,” Tambe explains. “In lower-risk areas, fewer snares were discovered.” However, the quality of data impacts PAWS’ predictions. As Shahrzad Gholami, a data scientist at Microsoft, highlighted, snares may go undetected if they are well hidden, and rangers can only gather limited information on poaching activities based on what is found in the field.

Buoyed by their success at QENP, Tambe’s team partnered with the World Wildlife Fund (WWF) in 2018 to bring PAWS to the Srepok Wildlife Sanctuary in Cambodia. This sanctuary, home to Asian elephants, bongos, and leopards, faces similar challenges to QENP, including vast terrain and few rangers. The PAWS team worked with local conservationists to refine their model, discovering key insights, such as heightened poaching risk near Route 76, a highway through the park. They also noted different poaching tactics used by Vietnamese and Cambodian poachers, particularly during the monsoon season.  During the PAWS field test, the team found five times more snares than average. 

The SMART system played a crucial role in these tests. Years of patrol data logged in SMART made Srepok an ideal field testing site for PAWS. The SMART Mobile platform, built on CyberTracker software, allows rangers to collect and log GPS-tagged data during patrols. New features include navigation tools, the ability to manually add GPS points, a statistics page for tracking patrol progress, and the option to add photos and audio to observations.

With further trials planned in Uganda and continued successes in Cambodia, PAWS and SMART are proving essential in making patrols more targeted and responsive to evolving poaching threats.

Drones: Patrolling the Skies

Drones, equipped with AI, have proven to be another valuable tool in anti-poaching efforts, offering a broader range of surveillance over large terrains. These unmanned aerial vehicles (UAVs) can fly over national parks, scanning for unusual human activity, and relaying real-time information back to rangers. The incorporation of thermal imaging helps detect poachers operating under the cover of night. High resolutions cameras beam footage and locations back to mobile control centres where ground crew are operating the drones. The cameras capture images that will help identify individuals, vehicles etc if there is enough light, but in low light Night Vision cameras will use infrared light to bright up the scene before taking photos.  AI algorithms help identify human vs animal activity in vast wildlife areas, making it easier for rangers to respond quickly.

While they may not be the ultimate solution, drones hold significant potential, especially in smaller reserves. Their mere presence can deter poachers. For instance, in Kruger National Park in South Africa, a six-week drone surveillance in 2014 led to zero rhino deaths in areas where poachers had killed nine rhinos just the month before.

‍

‍
Sadly Air Shepherd suspended wildlife poaching operations during Covid, but their drones were operational in Zimbabwe’s Hwange National Park, Malawi’s Liwonde National Park, and Nkhotakota Wildlife Reserve, among other locations. The impact of Air Shepherd’s drones were significant. In Hwange National Park, where cyanide poisoning of elephants at waterholes was rampant, the drones helped reduce poaching by an impressive 65% during the pilot project. There were no poaching deaths in areas where drones were actively used.

DJI Mavic 2 Enterprise Series are popular drones which have been used in Kenya and other parts of Africa to monitor and protect endangered species, including elephants and rhinos. They have also been used in Austrailia during the bushfire crisis 2019-20 where emergency crews used these drones to detect the heat signatures of koalas. After locating the animals, the drone’s 30x-zoom visual camera allowed responders to assess whether the koalas required medical care.

‍
‍

‍
‍

Despite these successes, drone technology faces limitations, such as the need for skilled operators and challenges in covering vast areas for extended periods. Nevertheless, they remain an essential part of a multi-faceted approach to wildlife protection.

Detecting Wildlife Snares with AI

Another emerging use of AI in wildlife conservation is in the detection of snares. Snares are traps set by poachers that indiscriminately harm a wide range of animals. AI-powered solutions like airborne synthetic aperture radar (SAR) technology have made it possible to mass-detect snares from the air. These radar systemsdeployed via drones or aircraft, use high-resolution imaging to identify snares hidden under dense foliage. Early trials of this technology in African reserves have shown promising results, dramatically increasing the detection rates of these deadly traps.

This AI-powered snare detection capability is vital, as traditional methods often miss snares, which are hard to spot even during patrols. By identifying snare locations more accurately, rangers can prioritise areas for removal, helping to save thousands of animals from a slow, painful death.

Combining Predictive AI and Real-Time Surveillance

AI systems like PAWS don't just rely on cameras or drones; they take a more comprehensive approach by predicting where poachers are likely to strike. By using machine learning to analyse vast amounts of data, including terrain, wildlife movements, and past poaching incidents, these systems generate predictive maps that highlight areas at high risk of poaching. This allows rangers to focus their efforts on vulnerable zones, making their patrols more efficient and potentially preventing poaching before it happens.

Other companies, such as AxxonSoft, use AI-driven video surveillance to monitor reserves and distinguish between human and animal activity. Their systems provide real-time alerts, enabling rangers to respond immediately to potential threats. AxxonSoft's technology has already been deployed in several African parks, where it has proven successful in reducing poaching.

The Future of AI in Anti-Poaching

As AI technology continues to evolve, its application in anti-poaching initiatives will only grow. Whether through predictive algorithms, fixed cameras, drones, or radar systems detecting snares, AI offers a powerful set of tools to combat poaching across diverse landscapes. The technology is helping conservationists make the most of limited resources, ensuring that wildlife protection efforts are more strategic and effective.

In conclusion, AI is playing a crucial role in the fight against poaching in Africa, Cambodia, and South America. By combining ground-based surveillance with aerial monitoring and predictive analytics, conservationists now have a range of AI tools to better protect endangered species from illegal poaching activities.

‍

‍

‍

‍

‍

‍

Website References

1. https://wildlabs.net/discussion/mass-detection-wildlife-snares-using-airborne-synthetic-radar

2. https://news.mongabay.com/2021/10/drones-are-a-knife-in-the-gunfight-against-poaching-but-theyre-leveling-up/amp/

3. https://www.ssass.co.za

4. https://www.africanews.com

5. https://seas.harvard.edu

6. https://www.africaoutlookmag.com

7. https://www.heliguy.com/blogs/posts/conservation-drones?srsltid=AfmBOoqAHrTlTOxtCmiEskU5Ru8negKCzahSkucOlk_SjLOH7i_q1G6H

8. https://en.reset.org/shepherding-rhinos-and-elephants-safety-drones-07262020/

9. https://www.ljmu.ac.uk/research/impact-hub/impact-videos/ai-cameras-enable-real-time-tracking-of-threatened-animal-species

10. https://www.prospects.ac.uk/podcasts/future-you-podcast-transcripts/how-artificial-intelligence-can-help-endangered-species-with-ljmu-podcast-transcript

‍