The DigiSchwein project contributes to the further development of animal-friendly, resource-efficient pig farming. In close cooperation with joint partners from industry and research, the project will develop solutions based on IOT, Big Data and Machine Learning to improve animal welfare, animal health and resource efficiency and reduce the input of nutrients into the environment.

Pig farmers are obliged to inspect their livestock at least twice a day. Farmers must rely on their experience and their subjective impression of the animal during their inspections. An increased noise level in the barn, wounds on ears or tails are alarm signals which the experienced farmer registers and acts accordingly. In contrast, the early detection of fever, the refusal to take in water or the development of abnormal behaviour are more difficult or impossible to detect.

To support farmers in their daily work, the DigiSchwein project will develop an early warning and decision support system. The focus of this work is on the solution of current practical problems in the farming of pigs. These use cases include the keeping of uncupied pigs and the associated prevention of tail biting, early detection of disease in order to isolate affected animals as early as possible, farrowing management to prevent crushing of newborn piglets and to minimize sow fixation times and monitoring of nutrient flows in the pig barn. Based on cameras, thermal imaging cameras, climate sensors that measure the ammonia concentration in the air, NIRS sensors to measure the nutrient composition of manure and other sensor technology, the conditions in the barn and the status of the animals are continuously monitored.

Not only the selection of suitable sensor technology that captures the relevant aspects for the applications is a challenge, but also the massive amount of data, which can grow to several 100 TB in the course of the project, mainly due to a multitude of more than 50 cameras. The storage and processing of the data is handled by a data management and data analysis platform developed with open source Big Data components. With the help of data stream processing, sensor data is checked for plausibility, merged and events are generated, such as exceeding a specified temperature in the barn. On the other hand, Deep Learning is used to develop models that allow the detection of events in videos, e.g. tail biting. The generated events then form the basis for the creation of specific data products which are integrated into the early warning system in order to provide recommendations for action. Through continuous exchange with farmers, the DigiSchwein project supports the transfer of the knowledge and results gained into broad agricultural practice.

• Data Stream Analysis
• Relational and NoSQL Databases
• Open Source Big Data Technologies
• Machine learning