|(Dezhen Song photo)|
On the trail of the ivory-billed woodpecker
Robotic cameras are searching for the elusive Holy Grail of birdwatching, and it makes a lot of sense: They're cheaper than humans, aren't afraid of snakes, and don't disturb the local environment
| 28 February 2007
In the bayous of eastern Arkansas, amidst ancient trees both living and dead that provide nourishment to creatures of the swamp, hangs a high-tech sentinel patiently waiting to capture video of an elusive bird once thought to be extinct.
Developed by researchers from Berkeley and Texas A&M University, the high-resolution intelligent robotic video system installed in the Bayou DeView area of Arkansas' Cache River National Wildlife Refuge is part of a major effort to locate the ivory-billed woodpecker in its historic habitat, the bottomland forests of the southeastern U.S.
If the researchers obtain conclusive photographic evidence of the woodpecker, it will settle a heated debate that has fascinated millions around the world, from birdwatchers and environmentalists to Arkansas farmers and duck hunters.
In the meantime, the new robotic video system provides detailed video sequences of other birds, suggesting a new high-tech approach to doing field-biology work.
The video system is part of a new project, called Collaborative Observatories for Natural Environments (CONE) and funded by the National Science Foundation, to develop automated systems that can observe and record detailed natural behavior in remote settings. The co-principal investigators of the project are Ken Goldberg, a Berkeley professor of industrial engineering and operations research and of electrical engineering and computer sciences, and his former graduate student, Dezhen Song, now an assistant professor of computer science at Texas A&M.
In the early stages of the video system's development, Goldberg worked with Berkeley's Robert Full, a professor of integrative biology. "We call it COLABCAM, for collaborative lab camera," says Full, "and it is meant to show experiments in progress on animals to our robot collaborators. What Goldberg has now done is take this lab camera into the field."
Goldberg and Song recently teamed up with researchers from the Lab of Ornithology at Cornell University to help look for the ivory-billed woodpecker, reverently referred to as the "Lord God bird" or the "Holy Grail of birdwatching." The bird, standing 18 to 20 inches tall with a wingspan of 30 to 31 inches, was considered the largest woodpecker north of Mexico when it vanished more than six decades ago. It feasts on beetle larvae that inhabit the dead or dying trees of the bottomland wilderness - or what remains of it - in the southeast United States.
It had been feared extinct for decades until sightings in recent years revived hopes of the species' survival. However, most eyewitness observations of the woodpecker were made while the bird was flying through dense forests, making it difficult to obtain photographic evidence of the sightings. Yet biologists made national news headlines in 2004 when they captured a few seconds of video of what appears to be an ivory-billed woodpecker.
But those sightings also generated their share of controversy, with skeptics claiming that the fuzzy image in the video actually depicted a common look-alike bird, the pileated woodpecker.
Hopes that the ivory-billed woodpecker survived extinction were renewed last September along the Florida Panhandle when another team of scientists, from Auburn and Windsor universities, reported sightings and retrieved a number of audio recordings of the bird's distinctive double knock. Still, none of the evidence to date has provided the definitive proof of the bird's existence that skeptics demand.
"A single photographic frame would have to clearly show the unique markings of the ivory-billed woodpecker," says Goldberg. "Much better would be a high-resolution video clip that would also capture its unique wing and flight patterns."
The researchers note that simply pointing video cameras at the sky and recording is not practical, as the images would quickly fill up the computer's hard drive. The challenge, they say, is for the software to automatically recognize when animals are present. "Passive infrared [PIR] motion sensors are sometimes used in wildlife research," says Goldberg, who has pioneered networked teleoperation systems for more than a decade. "The problem is that PIR sensors look for heat and are not triggered by birds flying overhead. So we're developing a robotic system that analyzes high-resolution video in real time."
A natural bottleneck
In February 2006 the Cornell researchers took Goldberg and Song out to the Cache River refuge to scout out potential locations for placing the remote cameras. Because no one knows exactly where the bird might appear, the cameras must cover a relatively wide swath of sky. They settled upon a power line that cuts through the bayou and provides a 50-foot-wide clearing unobstructed by trees.
"It's a natural bottleneck in the forest, and birds passing through that corridor are relatively easy to spot because they expose themselves," says Ron Rohrbaugh, project director at the Cornell Lab of Ornithology. "At this location we should have the highest probability of capturing an image of the ivory-billed woodpecker."
The two cameras - one pointing east and the other west - are connected to a computer that processes the data. Waterproof gear helps protect the equipment from the elements, including rain and wind, and even from occasional bird droppings.
The researchers created software that keeps video files only when potential "bird flight" movement is sensed. The software is based on new algorithms that can handle the unpredictable conditions of a natural environment, filtering out false readings from clouds, water reflections, and falling leaves. "The program knows, for instance, that the ivory-billed woodpecker flies 20 to 40 miles per hour, so anything outside that range is deleted," says Song, who helped develop the software. "The high-resolution camera we have shoots at 22 frames per second, with approximately 2 to 3 megapixels per frame. That's a huge amount of data that must be managed."
Collecting the video data involves a decidedly low-tech approach: David Luneau, associate professor of electronics at the University of Arkansas at Little Rock, takes a boat out to the site every two weeks to change the disk.
Not only is Luneau skilled with computer equipment, he is an avid birdwatcher and a leading member of the ivory-billed-woodpecker search team in Arkansas. He does an initial screening of the images from the hard drive, then sends the data to researchers at Cornell, Texas A&M, and Berkeley. If a high-quality image of the ivory-billed woodpecker is eventually captured, the ultimate call on the bird's identity will be made by Cornell, says Song.
Rohrbaugh pointed out the benefits of using an autonomous camera. "There are other ways of searching for the ivory-billed woodpecker, but those ways usually involve a human positioned in the forest for a very long time," he says. "Humans are expensive, they're not always alert, and their simple presence is a disturbance to the environment, even when they're camouflaged and sitting quietly. Remote systems that can serve as our eyes and ears are a big advantage."
Song also noted that using the camera extends the search season to the entire year.
"Usually people do this type of birdwatching in the winter because there are fewer leaves, making it easier to spot the woodpecker," Song says. "Also, in the summer, the temperature is hot, it's swampy, and there are mosquitoes and snakes to deal with. Our system can run the whole year, and it is not bothered by mosquitoes."
The researchers are continuing to fine-tune the system and algorithms while combing carefully through each new set of video that is collected.
"I'm a person who's been in the outdoors all my life, and I'm trained as a wildlife biologist," says Cornell's Rohrbaugh. "Certainly going into this I had a lot of skepticism about the usefulness of this robotic camera. But now there's hope that by using this camera we can get a high-resolution image that will show, indisputably, that the ivory-billed woodpecker is living in Arkansas."
However, the researchers also acknowledge the possibility that the robotic cameras may never capture definitive footage of the famed woodpecker.
"I'm hopeful, but not overconfident," says Goldberg. "We're willing to run this camera for years, and we're prepared to accept it if we never see the bird. But if this persistent robot out on the bayou manages to capture verifiable high-res images of the legendary ivory-bill, it would be a major discovery for scientists, for conservationists, and for more than 45 million American birdwatchers."