Remote Activity Classification of Hens Using Wireless Body Mounted Sensors

Abstract

This paper presents the design and implementation of a machine learning based activity classification mechanism for hens using a wearable sensor system. Legislation and social demands in the U.S. and Europe are pushing the poultry industry towards the usage of non-cage housing systems. However, non-cage systems typically house hens in groups of hundreds or thousands, which makes it nearly impossible for caretakers to visually assess the health, welfare, or movement of individual hens or to follow a particular hen over time. In the study, laying hens were fitted with a lightweight (10 g) wireless body-mounted sensor to remotely sample activity data. Specific machine learning mechanisms are used on the features extracted from activity data to identify a target set of activities of the hens. The paper establishes technological feasibility of using such body-mounted sensor systems for accurate hen activity monitoring in a non-cage housing system.