Digital image analyses as an alternative tool for chicken quality assessment

Poultry meat colour is an important quality attribute for the rapid detection of “pale poultry syndrome”, as it is affected by conditions of animal welfare during pre-mortem period. The meat processing industry demands a fast and non-contact method for accurate meat colour assessment. In the present study, computer vision was tested as a potential tool to predict colour measurements compared to CIELab attributes of chicken breast ( pectoralis major ) obtained by analytical reference measurements. The proposed approach using computer vision was successful in avoiding pixels with little information (specular reflection) and based on an illumination normalisation step it was obtained an acceptable correlation between colorimeter measurements and the proposed framework (Delta E  = 5.2). High correlation coefficients obtained between computer vision and colorimeter validate the approach for measuring L* colour component. Results for determination coefficient was R 2  = 0.99 for L*. In addition, our framework reach R 2  = 0.74 for a *, and R 2  = 0.88 for b * component. Results suggest that computer vision methods base on an RGB device can become useful tool for fast quality assessment of chicken meat in large-scale processing plants.