Using video image analysis to count hens in cages and reduce egg breakage on collection belts

Abstract

Stock people working in modern cage layer sheds spend more than half their daily work time directly checking hens to monitor health and welfare. In addition, mechanical egg collection belts must be checked for potential blockages that may result in cracked or broken eggs during the collection process. These are important tasks in the profitable management of modern multi-tier cage systems. However, where the upper tiers of cages are above stockperson eye level, the effectiveness of humans to perform these tasks accurately may be questioned. We investigated whether video image analysis (VIA, the ability of a computer to ‘see’) could automatically perform two common tasks – that of counting the number of hens per cage and scanning the egg collection belt to identify foreign (non-egg) objects. Cameras were attached to the robotic feeder that moved along the front of the cages. Views of the interior of the cages and the egg collection belt were recorded on digital video as the robotic feeder moved. Two VIA prototypes were evaluated, initially at the research institute and subsequently at a commercial farm. Using the respective automatic detection algorithms that were developed for the research, 79% of targets (hen legs) in cages were correctly counted, while 95% of foreign objects on the egg collection belt were detected. The results demonstrate that VIA can be used to monitor egg belts for potential blockages, and has potential as technology to count hens.