University of Nebraska-Lincoln (UNL) researchers are engineering new precision technology to help producers continuously monitor animals and use the data to ultimately improve animal well-being.
The system processes 24/7 video footage from livestock facilities and applies machine learning that uses statistical algorithms to help computer systems improve without being explicitly programmed, according to a university release. The technology identifies individual pigs and provides data about their daily activities, such as eating, drinking and movement.
“We want to make a tool that is available to the livestock producers,” said Ty Schmidt, UNL associate professor of animal science. “In a competitive agricultural market with rising costs, producers are looking for solutions that streamline operations while enhancing the health and well-being of their animals.”
The UNL interdisciplinary team includes electrical and computer engineers Lance C. Pérez, Eric Psota and Benny Mote, and Schmidt, who developed the technology using video footage of pigs.
So how does the technology help producers? Researchers say this system can estimate how much each pig weighs and its rate of growth.
“I believe we can make significant strides in the efficiency of pork production with this technology,” Schmidt said. “Ultimately, we’d like to see this system rapidly and accurately identify pigs that are sick or displaying destructive behavior prior to observations by caretakers. We are hopeful that the system will be able to assist producers throughout the entire production system.”
From a behavior standpoint, the team is working to program the system to identify animals that are displaying symptoms of illness and pigs exhibiting aggressive/damaging behaviors such as fighting and tail-biting. The technology can also help identify a pattern of typical behavior.
“When an animal deviates from that pattern, then it may be an indicator that something’s wrong. It makes it easier to spot problems before they get too big to fix,” said Psota, research assistant professor of electrical and computer engineering. “
Using deep learning networks, the team created a system utilizing a form of machine learning with millions of coefficients and parameters. To identify pigs from all angles, the networks processed images large and small, rotated, skewed and otherwise transformed. The team uses ear tags to help with identification but plans to rely on unique physical characteristics such as ear shape, saving producers the added work of tagging.
“We would also like the system to identify abnormal changes in behavior related to other factors, such as environmental changes like heat stress and cold stress, Schmidt added.
The team is pursuing further development with the help of NUtech Ventures, the university’s technology commercialization affiliate. Although the system has been developed to identify pigs, its algorithms can be used for other livestock, such as cattle, horses, goats and sheep.
In the future, the team plans to explore the technology’s ability to predict illness. They recently received $675,000 from the National Pork Board to fund two studies. In collaboration with Kansas State University, the team plans to collect data from both healthy and immune-compromised pigs, training the system to distinguish early symptoms.
The second study will explore the lifespan of sows — female pigs of reproductive age — and traits that may be associated with longevity. The Nebraska team’s technology will track sows over time and identify changes in movement, gait patterns and physical activity — data that may yield links between genetic background and longevity, the release said. It’s a connection that hasn’t been measured because there hasn’t previously been technology to do it, Schmidt said.
Farm Journal’s PORK | Jennifer Shike |