Field Validation of a Visible-Light Camera-Based Calving Detection System in Commercial Beef Cattle Farms

Efficient monitoring of cows during the periparturient period is essential for reducing calving-related losses and labor demands. This study evaluated a camera-based calving detection system using visible-light images under commercial farm conditions. A total of 61 cameras were installed on 31 beef cattle farms, and data from 357 calving events were analyzed. Behavioral parameters, including movement distance, posture changes and tail-raising behavior, were analyzed using an artificial intelligence-based system. An alert was generated when the estimated calving probability exceeded 70%. Using a predefined criterion of at least one alert within 24 h before calving, the system successfully detected 336 of 357 calving events, resulting in a sensitivity of 94.1%. A total of 448 successful alerts and 191 early alerts were recorded, resulting in a positive predictive value of 70.1%. The mean interval between the first alert and calving was 192 ± 9 min, and the median interval was 142 min (interquartile range: 75-250 min). Alert frequency increased significantly as calving approached (p < 0.05). These results indicate that visible camera-based monitoring can effectively detect calving events and may provide a practical, non-invasive tool for improving calving management in beef cattle.