Livestock

In livestock farming, we work with advanced computer vision methods to develop practical solutions for the diverse challenges faced by this agricultural sector. These methods are tailored to the individual requirements of different species, ranging from cows and horses to poultry and even fish and insects. The criteria that matter to stakeholders in the sector such as usability and price tolerance in relation to hardware are taken into account from the outset.

The technologies we have developed facilitate the constant and fully automated monitoring and documentation of animal behavior and vital data and are thus supportive in achieving compliance with and verification of animal welfare standards. In this way, we assist farm operators in meeting ecological and social standards as well as maximizing economic sustainability.

We employ computer vision technologies to tackle specific problems in the livestock industry. By applying image processing algorithms and machine learning, we can extract valuable information from the image and video material obtained and use it to optimize livestock management.

1. Behavior monitoring: Image data collected in the barn can serve for behavior monitoring by means of computer vision techniques. Such analyses include movement and feeding patterns and make it possible to differentiate between normal and abnormal behavior of the livestock that might indicate the onset or presence of disease.
2. Vital data monitoring: Non-invasive, regular and even continuous vital data monitoring of pulse rate and temperature is made possible by the analysis of image data. Such an approach offers enormous potential not only for the (early) detection of disease, but also in birth and reproduction management.
3. Early detection of disease: Data fusion between computer vision-based features such as behavioral and vital data monitoring form the basis for (early) disease detection. Merging this data over sufficiently long periods of time enables the creation of an individual medical record for each animal, which can also be used to identify complex disease progressions.
4. Animal welfare: Animal welfare can also be observed by combining information from behavioral and vital data monitoring, thereby ensuring and demonstrating compliance with animal welfare standards.
5. Barn hygiene: Multispectral and/or hyperspectral imaging is used to detect germ foci on surfaces in the barn and to identify the causes of isolated or even frequently occurring animal diseases.
6. Digital barn twin: The digital barn twin offers a virtual representation of a cowshed or stables. This can be used, for example, to illustrate a barn design or to simulate a real barn by feeding in real-time data.

The solutions we have developed give us detailed insights into animal behavior, health and nutrition. The automation of monitoring and analysis processes helps livestock farmers make better management decisions, improve animal health and welfare and increase the efficiency of their farming operations.

We are working with the Research Institute for Farm Animal Biology (FBN) at Dummerstorf to develop digital concepts for the barn. In current farming practice, extensive and systematic measurements and evaluations of animal welfare parameters in herd management are rarely carried out due to the high complexity of multisensory systems. In many cases, livestock farmers go no further than using conductivity measurement sensors.

Continuous recording of animal welfare parameters in different environments (i.e. barn or pasture) requires a high level of technical planning and configuration. Taking bovine welfare as an example, a sensor and data platform based on the plug & play principle is therefore being developed to close the gap between currently available solutions and practical application.