Engineers, wildlife biologists join forces for an inside view of life in the wild
Humanity’s interconnection to the animals that share our planet is an inevitable fact of life and, at the same time, an intricate mystery. We are endlessly fascinated with animals’ habits and antics, marveling both at what we have in common with them and the vast differences between us. The interplay between man and beast on a shrinking planet has become a topic of increasing concern.
That intimate relationship is the focus of a collaborative research project led by Henry He, an associate professor of computer and electrical engineering at the University of Missouri, funded by a pair of National Science Foundation grants.
“Since 2003, when I joined the ECE Department, my major research has been focused on integrated camera and sensor system design and their applications — eldercare, geospatial intelligence and force protection,” said He. “Now we’ll apply these technologies to wildlife monitoring and environmental changes where human observation is limited.”
He’s collaborators include two MU faculty members, electrical and computer engineering colleague, Assistant Professor Tony Han, and Professor Josh Millspaugh of fisheries and wildlife. Han’s mentor from the University of Illinois at Urbana-Champaign, Professor Thomas Huang, and Roland Kays, the curator of mammals for the New York State Museum, round out the far-flung team.
“We have put cameras and sensors together for a portable system to collect data about animals in the wild and analyze it,” said He. “These camera traps are attached to trees and animal motion triggers them. The system can last for weeks without interference.”
The wildlife biologists’ stake in the project is the accurate detection and assessment of animal populations and environmental changes, an undertaking that Millspaugh said usually involves complex fieldwork. The cameras make the undertaking much easier and also add accuracy and cost effectiveness to the task.
However, with hundreds of thousands of pictures, processing them becomes complicated, and this is where the engineers’ skill sets contribute to the project.
“We are employing computer vision, writing algorithms to analyze and classify the content, negating the need to manually view every frame,” said He. “We extract biometric parameters, such as body size, head size and speed of movement, as well as species ID.”
He said they eventually hope to identify group size, and in some cases, determine age and gender based on features like antlers.
“Using sensors, we will link images with temperature, habitat, and GPS weather data,” He added. “Right now we are building the tools and choosing the study sites.”
“It’s exciting to work with real data and also challenging,” said Huang. “With this collaboration, we can combine algorithms for better results.”
Image analysis tools won’t entirely take humans out of the equation.
“We’ll still manually look at selected images,” said Kays. “Human observation can help flag, for instance, if an algorithm thinks a possum is a skunk. But the algorithms will get better as we have human input. We started out pretty rough, but we are up to 80 percent accuracy on species ID.”
Kays also is working on screen measurements with He that will help identify animals’ distance from the camera and the direction of their movements.
Millspaugh and He previously collaborated with the Missouri Department of Conservation on a research project mounting tiny cameras to white-tailed deer, a species that is still a focus for the MU wildlife biologist.
“That was incredible work,” Millspaugh said of the earlier project. “There is a huge need for estimates of the abundance of deer in state parks and conservation areas. As large herbivores — when deer densities are too high — their presence has a great impact on things like plant communities.
“This research also gives us an opportunity to collect rare observations — insights we could not get any other way,” Millspaugh said.
Data gathered from the project will contribute to a Smithsonian searchable web portal making it accessible to both researchers and enthusiasts. And He’s team would like to maximize the amount of data collected by involving as many people as possible in gathering images and uploading them directly to the Smithsonian’s site.
“We’d eventually like tens of thousands of citizen scientists to work with us by giving them cameras and having them share their images with us,” Kays said, adding that this type of information already exists for birds because of coordinated efforts by birdwatchers.
“Plus, it’s fun and it gets people outdoors to learn about animals firsthand, connecting them with nature,” Kays added. “Deer hunters are already buying cameras, and we’d like to get them involved too.”
According to He, 60 cameras will be deployed in Missouri and rotated to different areas for a broader picture of life in the wild. He said they also are planning a system using a cell phone to collect data remotely from the cameras.
Many questions might be answered by mining such a vast body of data if it is managed correctly, such as how human actions impact wildlife and how different species coexist.
“If a neighborhood is built inside a big area of forest, some animals will be attracted to it,” said He. “Predators will follow if they depend on those animals, and other animals will move away. Human action will affect animal response for up to 20 miles away, but before, we didn’t have sufficient data to make assumptions.”
“Cross discipline collaboration between fisheries and wildlife and engineering makes this project really unique,” said Millspaugh. “It’s mutually beneficial and it has the potential to be transformational in our field.”