Showcase
It's like a Fitbit, but for cows
Nowadays, we want not only to be healthy, but healthier. And so are our standards for the health of our dairy cows. Many of us have watches giving us health advise depending on the biometrics it gathered. What if, we could develop a sensor for our cows allowing us to detect their behaviour changes as a health predictor? Obviously, cows cannot tell the farmer what is happening to them. Knowing more about their health status, and, detect their specific health issues as early as possible would not only benefit the cows themselves, but the entire dairy industry. Researchers at WUR, Claudia Kamphuis, animal breeding and genomics, and Marjaneh Taghavi, applied AI scientist, are analysing cow behaviour using sensors and AI in the Sensing Potential project.
How are they doing?
As Claudia Kamphuis says, a day in a cow’s life in the Netherlands is like a holiday inn with improvements to be made. They follow their natural behaviour of eating and ruminating in and out of the barn. They get milked two to three times a day, or as they please in cases where a milking robot is used. We could argue that they might spend time inside during the summertime, not eating fresh food during the winter months, but so are we. In any case, farmers do want their animals to be healthy, not only for their benefits and a nicer working environment, but also from the cow’s health perspective.
From situation to solution
The dairy industry has been working towards a higher efficiency, where the solution is often sought in larger herds. Consequently, this increase in herd size reduces the farmer’s time availability per cow. As a result, the farmers have relatively less time to observe their animals making it more difficult to detect injuries or infections. In the latter case, the earlier the infections are detected the less chance it spreads through the herd. The goal is to avoid a clinical situation, or shorten a clinical situation as much as possible with minimal effects.
The mother in me can only imagine the pressure of what is it to take care of hundreds of little beings, giving them personal attention while having hundreds more tasks to do in and out of the house, I mean the barn. Being unable to talk with them would make this task overly challenging. How do I know a cow might suffer before I notice a strong behaviour change making her stand out in the crowd? Of course, farmers know their cows. Farmers can quickly detect visually if there is something wrong with a cow. Plus, the current use of pedometer-like sensors helps detect if they are in heat and provide an indication that something is potentially ‘off’. In that case, the farmer must visually assess the cow, preferably multiple times a day, which is not always timewise possible. Until we learn how to speak the cow language or develop a system that could translate their thoughts and sounds, we must rely on their behaviour or physiological changes, such as milk quantities and quality. These changes are used as a tool to detect their health level using sensors.
Roger that?
The sensors studied during this project allowed to detect the personal behaviour change of a cow in terms of time spend doing their activities such as walking, eating, ruminating, and resting according to their surrounding environment. If the time spend for each behaviour changes for that particular cow it can be translated into usable information for the researchers. It was not done without any challenges. As Marjaneh explained: “We had very high expectations of the sensor.” However, the planned methodology had to be adapted due to the sensor not staying well in place on the cow. Furthermore, the lifespan of the battery could not gather months of data to convert into behaviour analysis. It had to be done with only a few weeks of data. Unfortunately, the sensors were battery dependent. During the charging time, another sensor would be put on that cow, not to lose data, and the newly charged sensor would be put back on another cow. The sensor management lead to a longer data management prior the analysis. To counter this challenge, a quality control had to be put into place to monitor the charging rotation of the sensors.
“The challenge is to get specific information about the animal. Not that the animal is hurt, but where, in what degree, and how it is compared to yesterday."
Blip, blip, new message from ‘cow’
The goal would be to achieve a diagnosis of the cow’s condition and not an approximation of a behaviour change that might be (in)directly linked to a specific health condition. Marjaneh is looking to improve the combination of different information source from camera (cow’s video footage), activity sensor and barn information with a AI learning process. The goal is to have more insight to the cow’s daily life and help the farmer to have a better view on his herd to better take even care of them, to know how the cows normally behave, notice the changes in behaviour and know what they need. The farmer could eventually get a notification on his/her smartphone about the specific health condition of a particular cow. Claudia: “The challenge is to get specific information about the animal. Not that the animal is hurt, but where, in what degree, and how it is compared to yesterday. The goal would be to develop a system that allows us to translate the cow language”.
The challenge is more to have better sensors, or to combined data from multiuple sensors to not say 'this animal may have this' but 'blip blip, this is saartje speaking, my toe is hurting, can you please have a look? It's worse than yesterday, it seems like I bruised my left front toe'.
Happily ever after
Let’s keep in mind that our high living standards allow us to have time and money to investigate and invest in solutions improving health detections of livestock which is not affordable everywhere. The Dutch dairy farming industry is doing as much as possible for the cows within the manageable 24-hour period for the farmer. This sensor could contribute to healthier cows, which also “explores the potential of nature to improve the quality of life”.