Where mist meets birdsong

Smart monitoring

In the smart monitoring center at Nandagang, publicity director Zhang Jingxing clicked a mouse, and real-time images from the wetland appeared on the large screen.

"Human patrols have limitations, but technology gives us eagle eyes and sharp ears," Zhang says, demonstrating the core functions of the integrated scientific research and monitoring platform.

"What we've built is an integrated monitoring network combining AI image recognition and soundprint recognition."

Zhang says that 11 sets of high-definition video monitoring equipment armed with AI are distributed at key places.

They are like eyes that never need to sleep, able to identify species and count the number of birds with an image recognition accuracy rate consistently above 90 percent, he adds.

"Even more powerful is our sharpears system," Zhang says, highlighting the three highly sensitive soundprint collection devices.

"Our soundprint database contains data on the calls of 1,569 bird species. It's like a bird sound dictionary," he says, displaying sound wave graphs.

According to Zhang, the system collects sounds in real time and automatically compares them with the species identification database.

So far, it has accumulated more than 220,000 valid soundprint records and has successfully identified 197 bird species.

The advantage of soundprint recognition is that it can penetrate dense reed beds, work in darkness, and monitor the birds without disturbing them, minimizing human encroachment, he says.

"For instance, on Aug 8, the platform, using soundprint recognition, successfully captured and confirmed the season's first southbound migrants — Temminck's Stints — before they were visible to the naked eye, recording 25 valid soundprints.

"This provides first-hand scientific data that previously could not be monitored, helping us accurately study migration times, routes, and population dynamics," Zhang says.