The imaging technology company is using AI and surveillance cameras to monitor pregnant cows 24/7, alerting farmers when calving is imminent
Imagine a second set of eyes on your pregnant cows and receiving an alert, with a high degree of precision and accuracy, when a cow is about to give birth. Now imagine those eyes trained on your pregnant cows 24 hours a day, seven days a week, taking neither breaks nor being called away to perform other tasks.
That is what Nikon is developing. The company, well known for its imaging technologies, is using its expertise to produce an advanced artificial intelligence (AI) system, using cameras to constantly monitor cows in individual calving pens. Employing advanced algorithms, the system can detect when a cow is about to give birth and immediately alert the farmer.
How the system works
“The system constantly monitors mother cows with multiple surveillance cameras and analyzes the cow’s characteristic behaviors by processing the images with AI,” says Hirotaka Kono, section manager with Nikon.
“Data is sent to the data center via the internet, and notifications are sent to producers’ and owners’ smartphones by messages and audio. AI detects the signs through machine learning, visualization of behavior trajectories, and segmentation machine learning content.”
Training AI
Kono says the biggest challenge in development was improving the ability to detect specific changes in cows’ behavior. “In the initial stages of development, the accuracy of detecting the characteristic behaviors of calving was low, resulting in many false notifications,” he acknowledges.
Performance has improved, he says, by collecting a lot of data and repeating AI learning, and by creating appropriate logic when generating alarms. He adds that the latest demonstrations show 95% detection accuracy for characteristic calving behavior.
AI training began in fall 2021, and the company continues to make improvements such as increasing the number of system-analyzed behavioral patterns.
Reducing burden on farmers
“Monitoring of mother cows requires observation of their condition day and night, which poses major challenges such as the physical burden on the farmer and securing human resources for farms,” Kono says. “In the future we believe that it may be possible to expand this system to monitor not only cows but also other livestock. We would like to contribute to reducing the burden on livestock farmers through AI and image analysis technology.”
Development has focused on 100-cow operations, common in Japan. But Kono says it will be possible to increase the number of cameras and develop new systems to recognize and distinguish large numbers of cows.
AI in livestock management
In agriculture it is becoming more common for computer programs to analyze mountains of data, increasing the number of more-informed management decisions. As more data is collected and analyzed, the result becomes even more useful.
“Dairy is the most advanced when it comes to using wearable collars and tags,” explains Paul Koffman, executive director of livestock technology solutions for Merck Animal Health. “Instead of someone doing heat detection two or three times a day on a dairy, we can track the cow’s activity and look at changes that would identify that she’s coming into heat.”
The power of predictive AI
The power of AI for livestock is that it can find relations between inputs and outputs that may not be readily obvious, says Tara Baker, North American marketing manager for Nedap.
“Everyone thinks that AI is a new revolution; in reality, we have been using AI for many years,” Baker says. “However, much of the use cases have been reacting to the data collected. The transition to predictive AI now is that we are using the data to look forward and provide more futuristic information to help producers succeed.”
Leveraging multiple sources
AI’s power also lies in its ability to compound vast amounts of information from several sources. “Nedap strongly believes in partnerships and integrations across the spectrum of technology suppliers to the farm that we are serving,” Baker says.
Opportunities abound to create a value picture at the farm level by fine-tuning available sensors or tapping into those of other industries, and using AI software. “It’s the power of being able to have a broader set of inputs to create stronger algorithms for a more robust look at the insights and provide accurate outputs,” she says.
