Evaluating pork grade by digital image processing

Left half carcass and loin eye pictures of 80 pigs were taken with a digital camera with fixed lens length and focus. After image processing, features were abstracted from the images. The correlative image features and the grades were used to train a Back Propagation Neural Network(BPNN) based on Digital Image Processing(DIP). Results indicate that fat thickness has significant relationship with image fat thickness(p<0.01). Carcass yield is correlative with image hunkers(p<0.01). Loin-eye area has a strong relationship with image loin-eye area(p<0.01). Muscle color is correlative with the mean 2G-B and the mean R+G of lean pixels in loin-eye region(p<0.01). Intramuscular fat characteristic is correlative with image intramuscular fat characteristic(p<0.01). Lean meat percentage was correlative with image fat thickness and image loin-eye area(p<0.01). In conclusion, the BPNN based on DIP can be used to evaluate pork grading quickly and accurately.