To Address the Problem of Human-tiger Coexistence, the First AI Camera That Can Identify and Transmit Tiger Photos Is Here
Since the beginning of the 20th century,The global wild tiger population has declined by more than 95%.It is listed as a red endangered species by the International Union for Conservation of Nature (IUCN). In recent decades, with the joint efforts of many countries and organizations, the number of wild tigers in the world has recovered, increasing from 3,200 in 2010 to 4,500 in 2022.
However,Another grim fact is the sharp rise in human population near tiger reserves.The expansion of human activities into tiger habitats has exacerbated poaching and human-tiger conflict. How to ensure the continued growth of tigers, a globally endangered species, while maintaining human livelihoods near tiger habitats has become a pressing issue.
Researchers from Clemson University have unveiled an AI camera alert system that can detect when a person is approaching a target.The AI algorithms it runs can detect tigers and poachers and transmit real-time images to reserve managers.
Compared with traditional camera traps technology, this system overcomes the former's shortcomings in false alarms and inability to respond quickly, and is expected to ease conflicts between humans and wildlife and promote progress in wildlife conservation.
Author | Tower
Editor | Sanyang, Xuecai
More than 100 years ago, there were about 100,000 tigers living in the wild.Today, there are fewer than 5,000 tigers in the wild.
Of the existing wild tigers, about 76% of them are distributed in South Asia, of which India has about 66% of the world's wild tiger population.The number of wild tigers in the country is close to the total number of all other tiger-distributed countries (Table 1).
a. Estimated tiger population outside designated tiger reserves in India;
b. Central India includes Madhya Pradesh, Chhattisgarh, Maharashtra and Telangana
Note: Central India is home to the world's largest tiger population, with the Kanha-Pench population being the largest in the region. Data collected between 2014 and 2020
An estimated 351 TP3T of India’s tiger population lives outside designated tiger reserves, and the species’ need to spread means this trend will continue to rise, with the result that tiger-occupied habitats will inevitably be highly disturbed by humans.
Tiger population recovery, species spread, and increased human population density around protected areas (Figure 1a)Both increase the risk of poaching and human-tiger conflict.
There are several main forms of human-tiger conflict:The resulting human and tiger casualties, including tigers killing livestock and attacking humans, as well as retaliatory killings of tigers by human residents, could reverse the growth trend of wild tigers and further threaten the support of residents living around protected areas for wild tiger conservation.
a. Between 2000 and 2020, the population around the world’s tiger reserves increased by more than 19.5 million people, with these population increases particularly affecting tiger reserves in northern and central India, Cambodia, and Sumatra;
b. The researchers initially deployed the TrailGuard AI camera alert system primarily in the forest corridor between the Kanha Tiger Reserve and the Pench Tiger Reserve in India, and also in the Dudhwa Reserve in northern India.
Given the vastness of tiger habitats and the limited presence of forest rangers,Provides real-time data notifications to conservation personnel,It will greatly help to quickly identify poaching suspects before or after poaching or conflict occurs, or minimize the negative impact of conflict.
Over the past twenty years,Camera traps are an important tool for biologists and natural resource workers to protect wildlife species.It provides a wealth of information on species density, distribution and timing, and plays a prominent role in monitoring and estimating the population of tigers and other cats.
However, traditional cameras have exposed many problems during use.These include false positives, high costs of collecting and processing data, and susceptibility to theft.All of this will undermine its effectiveness in wildlife conservation.
In addition, to this day,Camera traps still don’t have real-time notification capabilities.This is crucial to preventing poaching and human-wildlife conflict.
On September 25, 2023, researchers from Clemson University published a paper in the journal BioScience.Introduced an AI camera alert system called TrailGuard.The system can accurately distinguish tigers from other species and transmit relevant images to the terminal devices (mobile phones & computers) of the reserve managers within 1 minute.
Paper link:
https://www.science.org/doi/10.1126/science.adg7492
Follow the "HyperAI Super Neural" public account and reply with the keyword "tiger" to get the full PDF of the paper
TrailGuard AI Small size and easy to conceal
Compared with general camera traps,The camera portion of TrailGuard AI is much smaller.The version used in this study is about the size of a large pen (138 mm long, 14 mm wide, and 11 mm deep), with a camera connected to a communications device the size of a small notepad.
Traditional cameras are usually difficult to disguise on small paths close to the ground, but the compact TrailGuard AI can be deployed on trees together with communication equipment, or it can lurk a few meters away from the ground path.It has high security and confidentiality.
Left: TrailGuard AI compared to a quarter
Right: TrailGuard AI camouflaged near a trail in Dudhwa Tiger Reserve, able to send multiple real-time notifications of poachers without being detected.
Compared with traditional cameras, TrailGuard AI has the following advantages:
► Intelligent filtering of non-target species and false triggers,Only a small portion of the image captured by the camera is sent.
► Under the same charging conditions,Significant battery life savings are achieved because:
1. The cost of image transmission is much higher than the cost of loading and running inference on the AI module (this step is measured in hundreds of milliseconds), and embedded AI filtering can save a lot of image transmission costs;
2. When TrailGuard AI is not working, the entire system loses power.With the battery connected, the system can transfer an average of about 2,300 JPEG images and text files in a single pass, and can record about 1,500-2,000 untransmitted error trigger events on the SD card;
► Reduce false positive notifications and improve alert effectiveness.Traditional camera traps have too many false alarms, which may eventually cause users to habitually ignore alarm information. However, the detection alarms filtered by TrailGuard AI allow staff to determine whether there are potential poachers or wild animals and respond quickly.
In this study,TrailGuard AI is deployed in two locations (Figure 1b):
* A forest corridor between Kanha and Pench reserves in central India,The aim is to transmit the first batch of tiger images for pilot education targeting forestry departments and local residents;
* Dudhwa Sanctuary in northern India,The aim is to examine TrailGuard AI’s ability to provide real-time notifications when potential poachers are present and tigers are approaching villages.
System deployment and results Kanha-Pench corridor
The Pench Tiger Reserve in southwest India and the Kanha Tiger Reserve in northeast India have more than 300 tigers, the largest tiger population in central India and one of the largest in the world.
The Kanha-Pench wildlife corridor connecting the two protected areas covers an area of about 3,150 square kilometers and is a forest passage of about 140 kilometers in length (Figure 1b).
The tiger migration path involves hundreds of villages in the corridor and nearby areas. More than 2.7 million people live in the 5-kilometer buffer zone around the Kanha-Pench Tiger Reserve. The consequence of the convergence of people and tigers is thatHundreds of livestock are killed by tigers every year, and tigers also face threats from retaliatory killings by villagers and poachers.
From mid-May to mid-July 2022, the researchers simultaneously deployed 12 TrailGuard AI systems in the Kanha-Pench corridor over 591 trap nights.
The researchers worked with local rangers and guards to identify areas of recent tiger activity and cellphone data coverage, with all TrailGuard AIs located within a 1-kilometer radius of the forest-rural border, primarily on trails leading to villages.
In the study,All 12 TrailGuard AIs sent back images, and 8 of them detected the tiger (Figure 3). TrailGuard AI detected tigers with considerable accuracy in 61 trigger events—the median probability value for a marginal tiger detector was 0.9883.
TrailGuard AI detects a tiger within 30-42 seconds.Researchers and forest department staff were notified of the tiger’s presence, alerting everyone that tigers had been detected at three TrailGuard AI points within 300 meters of a village.
Additionally, researchers spotted a tiger at the site of a recent livestock kill, where it was foraging and subsequently moving livestock carcasses.
During the deployment, in addition to the tigers,Responsible personnel also received rapid notification of other species that are prone to conflict with humans.Including leopards, sloth bears and wild boars, the forest department immediately informed the village leaders via text messages.
System deployment and results Dudhwa Conservation Area
Dudhwa Tiger Reserve is a collection of protected areas covering approximately 1,310 square kilometers, which is home to approximately 107 tigers, as well as other species such as the endangered Indian rhino and Asian elephant. The high human density around Dudhwa makes the threat of poaching to tigers and their high prey populations particularly concerning.
For nearly 705 days from early September to mid-December 2022, researchers deployed seven TrailGuard AI alert systems in the Dudhwa Tiger Reserve.
During the study,All 7 devices successfully transmitted real-time images, and 3 of them captured tiger images.Forest staff and researchers receive tiger detection information every day from the two pilot deployment areas.
Among them, a detection point in Kishanpur Wildlife Sanctuary captured the same group of poachers carrying guns and knives on two dates. When the poachers appeared the second time, they had a body that was suspected to have been poached (Figure 5).
Three days after TrailGuard AI spotted the tiger, poachers reappeared at the same location.This shows that the threat of poaching faced by endangered wildlife populations is clear.Based on these transmitted images, authorities are able to quickly identify criminal gangs and launch investigations.
For decades, zoologists and wildlife biologists have longed for a smart camera alert system.Capable of detecting tiger and other endangered species activities and providing timely alerts,TrailGuard AI to some extent indicates that this dream has become a reality
TrailGuard AI with real-time notifications overcomes the drawbacks of traditional camera traps such as false alarms and inability to respond quickly.Transform wildlife protection from the previous passive use of camera data to the level of active early warning and response.
In protected areas with complex ecosystems, this technology can not only help protect villagers from attacks by predators such as tigers, but also prevent predators from being harmed by poachers. It also provides a good reference for other protected areas to deal with the relationship between humans and wildlife.
also,The researchers are considering expanding the application of TrailGuard AI.They set their sights on larger groups of animals, such as elephants:
"Elephants always wander outside human settlements, destroying crops, causing chaos in villages, and even causing train crashes, killing dozens of people every year.TrailGuard AI promises to prevent this from happening."
Reference Links:
[1]https://phys.org/news/2023-10-ai-tiger-tiny-camera-predatorand.html
[2]https://phys.org/news/2023-09-reveal-ai-based-camera-coexistence-tigers.html