The System For Evaluating Pork Carcass Grading Impersonally By Digital Image Processing Techniques

Pork Grading counts for much to the benefits of pig breeders and slaughterers. Visual evaluation and measuring manually were the existing assessing methods. These methods resulted in inefficient and errors associated with subjective evaluation. Consequently, pig breeders and slaughterers always disputed the truth of the results. Therefore, it was necessary to develop a system based on Digital Image Processing, which assessed pork grading accurately and automatically.Carcass and loin-eye pictures were obtained by Sony717 Digital Camera with fixed lens distance and focus. One hundred ternary hybridization pigs were chosen. Live pig weight, fat thickness, carcass weight, loin-eye area and lean meat weight were measured manually. The scores of meat color and intramuscular fat content were evaluated by the experts. Carcass yield and lean meat percentage were calculated based on the manual data. The pictures were processed by digital image processing. The features were picked up from the images and the models between manual data and image features were built. Then we evaluated pork grading by statistic analysis and Artificial Neural Network (ANN). The results are as below:(1) The criterion of pork carcass evaluation was improved. The new one evaluated pork grading by fat thickness, carcass yield, lean meat percentage, loin-eye area, the scores of meat color and intramuscular fat content.(2) With existing methods and improved region growth arithmetic in this issue, image features were obtained, including features of fat thickness, hunkers, loin-eye area, meat color and intramuscular fat content in the images.(3) The models between manual data and image features were established. The model between pork grading and relative image features was built, the equation was y=0.068x₁-0.011x₂-0.002x₃²-0.005x₄-0.011x₅-0.004x₆+4.903. F test and T test indicated the result was remarkable. (p<0.05, x₁ was image fat thickness, x₂ was the mean of G-B, x₃ was the ratio of white to red, x₄ was the mean of R+G, x₅ was hunkers, x₆ was the total number of pixels in loin-eye area).(4) ANN model was trained with treating relative image features as input and pork grading as output. The test showed that the model evaluated pork grading quickly and accurately and the system based on digital image processing run stably.The results indicated that the improved criterion could be altered by different target of meat processors and slaughters. It was more professional to evaluate pork grading, because besides carcass yield and lean meat percentage, other meat factors were included. The test showed that it was more reliable and had high precision to evaluate pork carcass grading by Digital Image Processing.