PubAg

Main content area

Predicting broiler gait scores from activity monitoring and flock data

Author:
Van Hertem, Tom, Norton, Tomas, Berckmans, Daniel, Vranken, Erik
Source:
Biosystems engineering 2018 v.173 pp. 93-102
ISSN:
1537-5110
Subject:
automation, birds, breeds, commercial farms, experts, farmers, flocks, gait, humans, lameness, models, monitoring, poultry, poultry industry, prediction, regression analysis, Europe
Abstract:
Leg deformities and lameness are a major welfare concern in the European poultry industry. This study was focused on the development of a prediction model for gait score in broilers, based on automated measures for flock distribution, bird activity levels and body mass. Data were gathered in five different farms in Europe. Broiler gait was assessed on a discrete [0–5]-scale by trained local human experts. Bird activity was continuously monitored with a camera-based system that automatically calculated the activity and distribution of the birds in the flock from the recorded images. Data analysis showed a linear trend between activity level of the flock on the day of the assessment (ACT) and the average gait score of the flock (GS): GS=−0.21∗ACT+2.85 (R2 = 0.55). Gait score and flock activity were negatively correlated (r = −0.741), whereas gait score and flock distribution was positively correlated (r = 0.705). Due to differences in management and broiler breeds, the absolute values in activity level and gait score vary between farms. The linear trend is however clear in all farms (R2 = [0.53–0.74]). Flock gait score could be predicted from continuous farm data by means of a linear regression model with a root mean squared error (RMSE) = 0.181 ± 0.003. This study shows that a camera-based monitoring tool for flock behaviour analysis has potential to warn the poultry farmer of possible gait problems in commercial farm settings.
Agid:
6122728