Thermal Imaging Efficacy in Predicting Trim Loss from Cull Sow Carcasses

Abstract

Abattoir-level digital image measurement and assessments recorded during the harvest process could contribute to better estimates of the economic value of each animal and could be used as animal health and well-being assessment tools to provide information/feedback for commercial pork producers. This study evaluated the association between live animal infrared images and nonedible trim loss from cull sows. Eighty sow carcasses selected from cull sows normally presented at a Midwest cull sow abattoir were evaluated. A thermal image and corresponding trim losses were collected from each sow.Regions of interest, referred to as hotspots, were isolated from the thermal images of individual sows. The relationships between the thermal hotspots, trim loss weight, and trim loss percentage from each carcass were evaluated using Pearson correlation coefficients, stepwise regression, and prediction equations. The most predictive model included covariates for the number of hotspots, live identification category, hotspot location, and explanatory continuous covariates for pixels with elevated temperature. Live animal defect classification and hotspot location categories significantly contributed to variation in percent trim. This suggests that visually identifying defects like abscesses, poor condition, mammary infection, or a combination on live cull sows is associated with trim losses from their carcass.