Mizzou Revolutionizes Bird Counting With Drones and AI

In the skies above Missouri's wetlands, a quiet revolution is taking place. Researchers from the University of Missouri are deploying advanced drone technology and artificial intelligence to redefine how we study and protect migratory water birds, paving the way for a new era in wildlife conservation.

These boxes are part of the advanced image analysis after the drone images are collected. Green boxes represent true bird predictions, yellow boxes represent false positive predictions and red boxes represent false negative predictions.

A Modern Solution to an Age-Old Problem

For decades, conservationists have relied on manned airplanes to conduct aerial surveys of bird populations. This traditional method, while essential, has always been loud, expensive, and carried inherent risks. The accuracy of these counts often depends on the sharp eye of a human observer, who can miss details as birds scatter or blend into the complex landscape below.

Now, a team at Mizzou’s College of Engineering, led by Professor Yi Shang, is pioneering a smarter, safer, and more efficient approach. By combining the aerial capabilities of drones with the analytical power of AI, they are creating a new system to track the migration patterns of key species like mallards and pintails.

How the Technology Works

The process begins with meticulous planning. The research team uses specialized software to map out the drone's flight path, optimizing for altitude, speed, and image overlap to ensure comprehensive coverage of the survey area. Once the drone captures high-resolution photos, AI algorithms take over.

These sophisticated algorithms analyze each image, identifying individual birds and carefully preventing any from being counted more than once. The AI can differentiate birds on open water from those in dense vegetation or fields. It can even be trained to identify different species, providing scientists with a far more detailed and nuanced view of the wetland ecosystem.

Unprecedented Accuracy and Efficiency

The results of this new method are impressive. "For straightforward situations, such as birds on open water, the technology is more than 95% accurate," explains Shang, a professor of electrical engineering and computer science. "Even in complex circumstances — where birds either overlap or are covered by trees or crops — our method is still 80-85% accurate."

To make this data immediately useful, the system integrates with large language models to analyze the image data and generate easy-to-read summaries. This provides the Missouri Department of Conservation and other agencies with actionable insights into the movements and populations of migrating birds.

The Future of Conservation Tech

This technology is not limited to counting ducks. As drones become more affordable and are equipped with even higher-resolution cameras, these detection methods will only improve. The same principles could be applied to monitor a wide range of other species and ecosystems, much like how AI is already used to analyze trail camera images of deer and other wildlife.

Through this innovative work, detailed in the study “New methods for waterfowl and habitat survey using AI and drone imagery” published in the journal Drones, Mizzou researchers are helping ensure that wildlife conservation can soar to new heights, protecting Missouri's migratory birds safely, accurately, and efficiently.