The aim of this study was to evaluate the use of thermal imaging as a rapid and non-destructive method for the detection of adulteration in minced beef. The adulterant materials investigated included avian offal (chicken gizzard) and ovine offal (sheep lung), which have been observed to be used as adulterants in the market. The weight ratios (w/w) of these impurities ranged from 0 to 30 % at 5 % intervals along with the pure samples. Since induction of a temperature gradient helps in thermographic diagnosis, heating and cooling pretreatments were utilized and compared. Convolutional neural network models were developed to identify the adulteration levels in ground beef. The models trained using data of both thermal processes (heating and cooling) were able to detect 8 levels of sheep lung adulteration in ground beef with 99 % accuracy. It was also able to detect the adulteration level of chicken gizzards in the warming method with 99 % accuracy and in the cooling method with 84 % accuracy. Since the method was capable of detecting impurities as low as 5 %, it points to a promising technique with high speed and accuracy for the detection of various types of adulteration in minced meat.
Read full abstract