Machine learning based fog computing assisted data-driven approach for early lameness detection in dairy cattle

Abstract Timely lameness detection is one of the major and costliest health problems in dairy cattle that farmers and practitioners haven't 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, we 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. The proposed approach has been validated on a real world smart dairy farm setup consisting of a dairy herd of 150 cows in Waterford, Ireland. Using long-range pedometers specifically designed for use in dairy cattle, we 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 behavioral activities, which are further analyzed in the cloud. Our hybrid clustering and classification model identifies each cow as either Active, Normal or Dormant, and further, Lame or Non-Lame. The detected lameness anomalies are further sent to farmer's mobile device by way of push notifications. The results indicate that we 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. Moreover, with fog based computational assistance in the setup, we see an 84% reduction in amount of data transferred to the cloud as compared to the conventional cloud based approach.