Early lameness is a considerable problem in the dairy industry. It causes pain and discomfort for the cow, while lowering fertility and milk yield for the farmer. Since current solutions come with high-initial costs and complex equipment, this use case utilises leg mounted sensors - measuring step count, lying time and swaps per hour - in combination with machine learning algorithms to identify lame cattle at an early stage. The collected data is analysed in the cloud and detected instances of lameness are sent to farmers’ mobile device to help treat affected animals immediately and avoid further effects. As opposed to a general approach, this use case customises the machine learning models to dynamically adjust as seasonal and herd conditions change. By detecting early lameness before it can be visually captured, treatment costs are decreased while animal welfare is improved.
Importance of data-driven detection
Timely lameness detection is one of the major and costliest health problems in dairy cattle that farmers and practitioners have not yet solved adequately.
The primary reason behind this is the high initial setup costs, complex equipment and lack of multi-vendor interoperability in currently available solutions. On the other hand, human observation-based solutions relying on visual inspections are prone to late detection with possible human error and are not scalable. This poses a concern with increasing herd sizes, as prolonged or undetected lameness severely compromises cows' health and welfare, and ultimately affects the milk productivity of the farm. To tackle this, Paul Malone, the Use Case Coordinator, and his team have developed an end-to-end IoT application that leverages advanced machine learning and data analytics techniques to monitor the cattle in real-time and identify lame cattle at an early stage.
Reproduction efficiency index
Their approach has been previously validated on a real world smart dairy farm setup consisting of a dairy herd of 150 cows in Waterford, Ireland. A farm where each cow has its own name which resembles how important it is for the farmer to solve an emerging health issue of the animals as early as possible. For this reason, the use case team uses long-range pedometers - specifically designed for use in dairy cattle – to monitor the activity of each cow in the herd. The accelerometric data from these sensors is aggregated at the fog node to form a time series of behavioural activities, which are further analysed in the cloud. Fog computing has emerged as a promising technology that can bring the cloud applications closer to the physical IoT devices at the network edge. A hybrid clustering and classification model identifies each cow as either active, normal or dormant, and further, lame or non-lame. Ultimately, the detected lameness instances are sent to farmer's mobile device by way of push notifications. The results indicate that the researchers can detect lameness 3 days before it can be visually captured by the farmer with an overall accuracy of 87%. This means that the animal can either be isolated or treated immediately to avoid any further effects of lameness and reduce significantly the need for antibiotics. Moreover, with fog based computational assistance in the setup, they observed an 84% reduction in amount of data transferred to the cloud as compared to the conventional cloud-based approach.
Challenges of exploring new testbeds
After the validation on a dairy farm in Waterford, Ireland, the use case integrated their infrastructure into the IoF2020 reference architecture and thus made interoperability points available to other third-party services running on this reference architecture.
The vendor dashboard application utilises FIWARE Orion Context broker and presents data using NGSIv2 compliant data models. By achieving this objective, the use case could access new testbeds in other regions. The use case is now deployed with two vendors using different sensor technologies and different architectures. To ensure greater flexibility for the farmers and end-users, the developed algorithms prove to work with both agri-tech solution providers. While ENGS uses the initial pedometer (leg-mounted) technology used in Waterford, Ireland, the new vendor, Herdsy utilises a collar-based system for herd tracking.
After the expansion to new regions, the use case now operates on five test farms in Ireland, Portugal and Israel with approximately 1400 cattle in total, on two vendor IoT platforms which all rely on the same machine learning approach. To further validate the solution, the team of researchers analysed pedometers data from another IoF2020 dairy use case: Herdsman+. Proving the interoperability and seamless integration, the primary goal of this validation, was successful. The resultant lameness detection performance was, however, inconclusive wholly due to the different characteristics of the provided data types. For the second objective to be successful, the researchers would have required a more extensive and targeted dataset from the other use case to detect lameness with their usual accuracy of 87% and 3 days before the farmer can visually detect it.
Days before visual detection
The end-to-end IoT application system with fog assistance and cloud support achieves this accuracy by engaging all its different farms and end-users alike. Also, the mobile app used by the farmers provides feedback from the farm and is used to retrain the machine learning models per herd. This engagement enabled this use case to provide richer and more accurate machine learning models and contributed to a selection of new features in the latest versions of the mobile app. Thanks to multiple language translations, the app gives user-friendly and actionable insights to improve farming practices, thereby increasing efficiency and yield. Dairy farms have all the constraints of a modern business - they have a finite production capacity, a herd to manage, expensive farm labour, and other varied farm-related processes to take care of. In this technology-driven era farmers look for assistance from smart solutions to increase profitability and to help manage their farms well. Through secure authentication and feedback mechanism, the use case solution aids the machine learning model re-training, thereby further improving the accuracy. Last but not least, the team also ensures that the collected data is only accessible by the specific end-user to whom it is related.
Achievements, products & services
Early lameness detection before it can be visually captured
Customised machine learning models to dynamically adjust to specific herd and environmental conditions
Improved animal welfare due to reduced antibiotic usage
Increased milk yield, decreased treatment costs