Analysis of feeding and drinking behaviors of group-reared broilers via image processing

Abstract

Farm managements and system designs could be improved based on the responses of broiler feeding and drinking behaviors. The objective of this study was to develop and validate image processing algorithms for automatic monitoring of feeding and drinking behaviors of group-reared broilers. Sixty Ross® 708 broilers at 26–28 days of age were kept in a 2.9 m × 1.4 m pen with a tube feeder and five nipple drinkers. Broiler behaviors in the pen were recorded and stored in images. Areas of concern near the feeder and drinkers in the images were segmented for broiler-representing pixels which were quantified to estimate bird number at feeder (BNF) and at drinkers (BND). Two days of data (24000 images) were used for algorithm training and testing. The results show that the algorithms had an accuracy of 89–93% for determining BNF. The mean square error between the predicted BNF and real BNF was 0.3-0.4 bird, indicating a good estimation precision of BNF by the algorithm. The sensitivity, specificity, and accuracy of the algorithms for determining BND were, respectively, 87–90%, 97–98%, and 93–95%. For most of the time on the sampling days, the feeder was occupied by 7–13 broilers simultaneously and each drinker by 0–1 broiler. Broilers showed spatial and temporal preferences in feeding and drinking, with more birds eating and drinking in areas with less disturbances, within a few hours after light ON and before light OFF, and during flock inspection periods. It is concluded that the algorithms had acceptable accuracies in determining BNF and BND, thus being useful components for vision-based behavioral monitoring systems.