Nature Lab

Waterfowl Management

Counting Waterfowl with Drones and AI

Counting ducks and geese is not always as simple as it sounds.

“Try to count 100,000 ducks from the ground when you’re at the same level as they are,” explains MDC Waterfowl Biologist Andy Raedeke. “The ducks can be packed together so densely that basically they look like a big, black blob. And if they’re in vegetation like standing corn, there’s no way to get an accurate estimate.”

Counts done from the air also have drawbacks. “If you’re flying, you’re going at 100 miles an hour, so you’ve got only seconds to come up with your estimate,” says Raedeke.

About five years ago, Raedeke began working with the University of Missouri and the U.S. Geological Survey on an alternate survey method: a drone with a camera and artificial intelligence (AI).

Reid Viegut, a graduate student, has been determining the best transect pathways for the drone to fly and the best height to take photos without disturbing the birds.

Computer science graduate students developed a mobile app to control the drone’s flight path and simplify camera settings, so the user won’t have to know photography to get a good image. They then created a desktop app to count the waterfowl in the photos. The app uses an AI model the students trained to distinguish between vegetation and birds. It even identifies the species (mallard, pintail, Canada goose, etc.).

This fall, Viegut is doing a full run and fine-tuning the process.

“I’m hopeful,” Raedeke says, “that as soon as next year, some MDC managers will begin doing their counts with drones.”

At a Glance

MDC and partners are developing a more efficient way to count ducks and geese by using drones and AI.

Partners

• USGS
• University of Missouri