With the opening of the new Thomson Nature Park, Singapore, the National Parks Board (NParks) recently unveiled a new system to help detect wildlife crossing the roads between forests and provide real-time warning messages for approaching vehicles, so that the vehicles can slow down and let the animals safely walk onto the other side (The Strait Times).
IronYun AI NVR Animal Detection solution
IronYun AI NVR Animal Detection is the solution that NParks has selected to realize the system
IronYun AI NVR Animal Detection is the solution that NParks has selected to realize the system. On one hand, AI NVR uses deep learning AI models to accurately recognize animals versus vehicles and people via camera feeds on the roadside. On the other hand, AI NVR integrates with signage systems to trigger the appropriate alerts when such animals are detected.
Wildlife protection program:
“Thomson Nature Park (TNP) is a 50-hectar (124-acre) green space to buffer between the eastern forests of the Central Catchment Nature Reserve and the new infrastructure developments. TNP is separated from the Nature Reserve by a 3-km-long Old Upper Thomson Road, on which visitor cars frequently travel. Animals, however, do not recognize manmade geographical boundaries”, said NParks Director for Conservation, Sharon Chan, and thus often walk across the road from the forest into TNP and vice versa to forage for foods and find mates.
Animal-vehicle collisions have occurred because the cars and motorcycles cannot see the animals and stop in time. Meanwhile, biodiversity surveys have indicated that many native animals, including critically endangered species such as the Raffles' banded langur and the Sunda pangolin often cross this road.
Several measures have been implemented to protect the animals, including:
- Aerial crossing: a rope ladder and a single rope crossing along the Old Upper Thomson Road to help canopy-dwelling animals crossing overhead
- Culverts: five culverts to help ground-dwelling mammals crossing underground
- Reducing traffic: turning Old Upper Thomson Road from a dual-lane road to a single-lane road in June 2018; plans to close the road to vehicles between 7:30 pm – 6 am daily in the future
- Roadway animal detection system: a combination of IronYun AI NVR, cameras, and signage to alert vehicles to slow down from afar when an animal crosses the road
In particular, the roadway animal detection system is co-funded by NParks and the Land Transport Authority (LTA) to reduce animal-vehicle accidents. At the heart of the system is IronYun AI NVR.
AI technology for wildlife protection
The system is guaranteed to recognize a human versus an animal versus a vehicle
AI NVR is a deep learning video analytics solution, which can distinguish several types of objects, including vehicles (car, bus, motorcycle, etc.), people, manmade objects (backpack, suitcase, etc.), and animals. IronYun engineering teams train the AI models using thousands of hours of video data to ensure high accuracy, so the benefits are two-fold:
- No false alarms: the system is guaranteed to recognize a human versus an animal versus a vehicle. As an improvement compared to legacy sensor-based systems, motions caused by tree branches swaying, people walking/biking, cars driving by, etc. do not trigger any alarm. In this case, only an animal crossing the road would trigger an alarm.
- Easy to use: no calibration to the environment is required. The model is pre-trained and ready to use from day 1.
AI Network Video Recorders (NVRs)
The LTA and NParks users set the alert rule so that when an animal appears, AI NVR recognizes, records the metadata and triggers lights to flash under a sign that reads ‘Animals Ahead’, all within 3 seconds. The car sees the flashing light and slows down, allowing the animals to reach safety.
The unintrusive monitoring and alert system AI NVR has proven useful
The current system supports 5 cameras along Old Upper Thomson Road. The system was announced on October 11, 2019, and is a year-long pilot project in the joint effort of LTA and NParks to protect wildlife in Singapore national parks.
Unobtrusive and alert video system
While the rope crossing and culverts help providing the animals safe alternative travel routes, Dr. Andie Ang, a primate scientist and chair of the Raffles’ Banded Langur Working Group, has commented that it would take time for animals like the Langurs to get used to artificial structures, so long-term monitoring is necessary.
Therefore, the unintrusive monitoring and alert system AI NVR has proven useful. According to LTA Chief Executive, Ngien Hoon Ping, “Joint efforts, such as the one on the roadway animal detection system, help us understand how technology could be deployed to achieve our aims.”