Precision Poultry Systems – Stress & Disease Detection via Sound

Quantifying the Effect of an Acute Stressor in Laying Hens using Thermographic Imaging and Vocalisations

Paper (Link)

The laying hen sector has multiple issues concerning the animal’s welfare. One crucial factor negatively impacts chicken welfare is stress. The conventional way of measuring and assessing chicken stress is time-consuming and subjective to the assessor. On the other hand, sensing and sensor technologies can be used to obtain objective, continuous and non-invasive/contactless measures of animal behavioural and physiological welfare indicators. The present study aims to investigate the use of thermographic imaging and microphones (sound) in obtaining objective indicators for acute stress in laying hens. During this study, 40 laying hens were stressed by opening an umbrella as a stressor starting from one day age until nine-weeks old. The birds were stressed every other day. Another 12 birds were housed in another cage. These birds were not stressed and served as a control group. The surface temperatures of the bird’s comb and beak decreased (1°C and 2.5°C respectively) in response to the applied stressor. This effect was only seen in the treatment group and not in the control birds.

The number of vocalisations the birds produced significantly decreased shortly after stress. The number of calls in the stressed group decreased from 39.5 to 12.1 calls/minute, where the control group decreased from 27.8 to 22.5 calls/minute. It was hypothesized that the number of vocalisations would increase after stress. This difference could be due to the daily behavioural rhythm performed by the birds. The birds might naturally produce more calls at certain hours of the day because of certain behaviours they perform. Three different neural network algorithms were employed to differentiate between the vocalisation of stressed and the control group. This was done by converting the audio files to images and feeding them to the pretrained convolutional neural networks (CNN). The Resnet CNN had the highest categorising accuracy with an overall accuracy of 86 percent. The changes in surface temperature of the beak, comb, eye, and head, as well as the results from the audio analysis could serve as potential indicator for acute stress in laying hens. Future research is warranted to validate the methodologies and findings under different environmental conditions and stressors.