Math vs. wild: Researchers use computer vision to identify wildlife
April Dahlquist, Coordinated Science Lab
- ECE Professor Thomas Huang will lead an NSF-funded research program on monitoring wildlife.
- To collect the data, motion-sensitive cameras will be set up in Missouri and Panama.
- This type of research is important to biologists who study the impact of human activity and environmental changes on wildlife species.
In June, ECE Professor Thomas S Huang received a three-year collaborative research grant entitled, “ABI Innovation Computational and Informatics Tools for Supporting Collaborative Wildlife Monitoring and Research.” The project, funded through a $443,405 National Science Foundation grant, includes collaborators from New York State Museum and the University of Missouri.
The research is looking to use information technology such as image processing and multimedia to monitor the movement and activity of wildlife animals. To collect the data, motion-sensitive cameras will be set up in Missouri and Panama to observe wildlife.
The grant’s aim is to create an algorithm that can identify the animal in view from its size, texture, etc. This way, analysis of the animal’s movement will be an automated process.
“Anything we can do to help make the task easier will be very important,” said Huang, the William L. Everitt Professor of Electrical Engineering and a researcher at the Coordinated Science Lab and the Beckman Institute. “We want to be able to automate as many tasks as we can.”
This type of research is important to biologists who study the impact of human activity and environmental changes on wildlife species. Scientists are hoping to advance and automate data collection and analysis when wildlife is present, identifying and classifying the animal, summarizing the scene and predicting animal geographic distributions.
The University of Missouri has already done some preliminary data collection, which shows encouraging results, Huang said. However, the challenge for Huang will be developing algorithms which work in real-life environments and not just in the lab.
“Anytime you try to apply computer vision to real cases you will see variations in data you did not anticipate in the development phase,” Huang said.
Since the cameras are triggered whenever they detect motion, the researchers have little control of the data collected, making this research challenging, Huang said. However, Huang is hopeful that the team will be able to create useful technology.
"Any techniques that make the task of biologists easier or faster will be very valuable," Huang said.