TY - JOUR

T1 - Applications of computer vision systems for meat safety assurance in abattoirs

T2 - A systematic review

AU - Sandberg, Marianne

AU - Ghidini, Sergio

AU - Alban, Lis

AU - Capobianco Dondona, Andrea

AU - Blagojevic, Bojan

AU - Bouwknegt, Martijn

AU - Lipman, Len

AU - Seidelin Dam, Jeppe

AU - Nastasijevic, Ivan

AU - Antic, Dragan

N1 - Publisher Copyright: © 2023 The Author(s)

PY - 2023

Y1 - 2023

N2 - Introduction in 2017–2019 of the new EU legislation on official controls in food production allowed use of computer vision systems (CVSs) as complementary tools in meat inspection of bovines, pigs and poultry. A systematic literature review was performed to identify and analyse relevant articles reporting on the performances of CVSs used in abattoirs for ante- and post-mortem veterinary inspection and meat safety assurance, including systems for detecting carcass/organ contamination and lesions. In this review, 62 articles were identified and analysed. There were 35 articles reporting on CVS performance in the detection of carcass/organ lesions and 27 in the detection of carcass contamination. CVSs for broiler chicken, pig and bovine meat safety assurance were reported in 53, 5 and 4 articles, respectively. Not all developed CVSs were validated, and only three articles reported results from real-time evaluation of CVS performance in an abattoir vs performance of the official veterinarian. Most of the reported CVS performance measures (i.e., sensitivity and specificity) were >80%. A high specificity in detecting lesions and carcass contamination (i.e., a low number of false positives) is of importance for the food business operator in order to minimise food waste, whereas a high sensitivity (i.e., a low number of false negatives) is required for production of wholesome and safe meat. At present, the existing CVSs developed for overall meat safety assurance of broiler chicken carcasses and organs demonstrate very high sensitivities but suboptimal specificities, indicating the need for further CVS development and optimisation.

AB - Introduction in 2017–2019 of the new EU legislation on official controls in food production allowed use of computer vision systems (CVSs) as complementary tools in meat inspection of bovines, pigs and poultry. A systematic literature review was performed to identify and analyse relevant articles reporting on the performances of CVSs used in abattoirs for ante- and post-mortem veterinary inspection and meat safety assurance, including systems for detecting carcass/organ contamination and lesions. In this review, 62 articles were identified and analysed. There were 35 articles reporting on CVS performance in the detection of carcass/organ lesions and 27 in the detection of carcass contamination. CVSs for broiler chicken, pig and bovine meat safety assurance were reported in 53, 5 and 4 articles, respectively. Not all developed CVSs were validated, and only three articles reported results from real-time evaluation of CVS performance in an abattoir vs performance of the official veterinarian. Most of the reported CVS performance measures (i.e., sensitivity and specificity) were >80%. A high specificity in detecting lesions and carcass contamination (i.e., a low number of false positives) is of importance for the food business operator in order to minimise food waste, whereas a high sensitivity (i.e., a low number of false negatives) is required for production of wholesome and safe meat. At present, the existing CVSs developed for overall meat safety assurance of broiler chicken carcasses and organs demonstrate very high sensitivities but suboptimal specificities, indicating the need for further CVS development and optimisation.

KW - Carcass contamination

KW - Computer vision

KW - Imaging

KW - Lesions

KW - Meat inspection

KW - Meat safety assurance

U2 - 10.1016/j.foodcont.2023.109768

DO - 10.1016/j.foodcont.2023.109768

M3 - Journal article

AN - SCOPUS:85151301110

VL - 150

JO - Food Control

JF - Food Control

SN - 0956-7135

M1 - 109768

ER -