"Cow tracking"

AI in the cowshed to make farmers’ work easier

The advantage of "cow tracking", especially for part-time farmers, is that they can keep an eye on their cows even when they are not on site.

(Bild: Jürgen Radspieler)

In future, farmers will be able to use artificial intelligence to find out about the current health status of their cows via an app on their cell phone. Automated "cow tracking" is currently being developed as part of a research project in Salzburg. 

Video cameras installed in the barn will transmit the data live to a software program. The system recognizes each individual cow and sounds an alarm if it is sick or injured or gives birth to a calf. Cow tracking has the advantage, especially for part-time farmers, that they can keep an eye on their cows even when they are not on site.

Ten farms in the Saalfelden im Pinzgau region are currently participating in the project run by the two Salzburg-based companies Mechatronik Austria and Cognify. The Austrian Research Promotion Agency (FFG) is funding the project with one million euros. For technical support, A1 Telekom-Austria is providing 5G equipment free of charge. The modems are a basic requirement for the fast transmission of live video data from the stables to the cloud. Soon there will be 20 farms taking part.

Four video cameras have been installed in the barn of farmer Michael Fürstauer's Vorderkühbühelhof farm. He can see the live images of his animals on his cell phone. After the planned completion of the project in 2025, he would immediately receive a message in the event of any abnormalities. If a cow is sick, ready to mate or about to calve, he can call the vet straight away. He can also use the live images to detect damage in the barn, for example if the drinking trough is no longer working.

Many tasks, one made easier
Fürstauer is married, has three children, is a truck driver and is therefore not always present on his farm. Another professional task of the Fürstauer family is renting out rooms on the farm. The farm will soon be certified as a "farm vacation". "I have 20 dairy cows and their offspring, ten calves. Cow tracking would help me with my work. For example, with oestrus behavior. The period for insemination is limited. If you miss it, it takes 21 days for the cow to get pregnant again. This is a major challenge with seasonal calving in conjunction with alpine farming. The information via the app that the cow is now in heat means that I can plan better and is also more economical," explains Fürstauer.

If the right time for insemination is missed, this means a financial loss for him of around 1500 euros per cow per year. "You also lose out on the additional funding for a new animal." When the app is fully developed, it will also have a lying alarm. If the cow does not get up after calving, possibly due to a lack of calcium, she could die the same day without appropriate treatment. The farmer is also informed immediately if there is a fire or disturbance in the barn.

Markus Zehentner, owner of the engineering and software company Mechatronik Austria and initiator of this project, describes calving as an emotional moment for a farmer in the small-scale sector. "The farmer is nervous, he wants to be there for the birth." Zehentner speaks from experience. He helped out on his relatives' farm in Saalfelden and trained as a skilled agricultural worker before studying mechanical engineering and founding the company.

During the coronavirus pandemic, he and his technicians carried out digital tests for a camera-based solution to support the farmer in his work. An acquaintance put him in touch with the data science service provider Cognify, which has been helping to develop cow tracking since 2023. "We learn from each other. A total of 13 employees are now working on the project."

The data from the cowshed is uploaded to the cloud via 5G, where it can be accessed at any time via an app.

(Bild: kuhtracking.com)

Animals always under observation
The system keeps an eye on the animals 24 hours a day. For this digital herd management to work and for each individual cow in a herd to be identified, a lot of data has to be fed in and evaluated using artificial intelligence. "The data is streamed to a cloud in real time. 20,000 photos are compared and matched again and again. We are now in the data collection and identification phase and are already able to recognize the individual cows," explains Zehentner. "Now comes the analysis work in relation to movement in the barn and the activity of the cows."

Mechatronik Austria is responsible for the technical infrastructure, the backend and programming the web interface on the cell phone, while Cognify is responsible for data evaluation, analysis and algorithms. "We train with the data from our AI models. Many different test cases are necessary for these machine learning models," explains Norbert Walchhofer, co-founder of Cognify. Some of the algorithmic challenges have been completed and the AI model chain is ready. "Object recognition and tracking are working, now we are in the re-identification phase." What still needs to be done is the classification of activities and health.

In addition to sustainability, the focus is also on animal welfare. There is no hardware on the animal. "Our solution is based on computer vision and therefore does not require expensive sensors on the animal," explains Walchhofer. The aim of such a system is improved economic efficiency, better husbandry conditions and improved animal health. "We are looking for sensible areas of application for AI. Cow tracking supports the farmer, he has fewer worries about his animals and it helps to maintain small-scale farming," says the expert, listing the benefits. In the future, this data will be used to feed a comprehensive herd management system that will make it possible to quantify animal welfare.

Data protection was not forgotten
Data protection was also taken into account. "A data protection expert has been involved in the project right from the start. The video data is protected accordingly," explains Walchhofer. "Only the farmer himself has live access." Signs in the test barns indicate that video recording is taking place. 

Once the project has been completed, a separate company is to be founded to distribute and sell the app. Investors are being sought for this. There is no concrete pricing model yet regarding the costs for the farmer. However, the annual fee should be relatively low.