Application Of Image Processing Techniques On Synthesis Evaluation Of Living Pig

Currently, the pig is graded mainly on the line of slaughter at home. The grades are given by evaluating the lean meat percentage in midsection of the pig in the traditional methods. This method has a low accuracy. Consequently, it may bring some economic loss to pigs feeder and process corporations.For solving the problems mentioned above, in this paper the study on the shape parameters of living pig was carried out to found a new method of assessing the grade of living pig. First, the shape parameters of living pigs were measured practically, such as the length, width and height of the body, chest depth, the length of one leg, the length of the abdomen. Then, the parameters of pork relating to the grades also were measured, such as back fat thickness, slaughter percentage and lean meat percentage. The living pigs 3-dimension images were took by the digital camera. Combined with image processing and pattern recognition, the shape feature parameters in image were extracted and then the grades of living pig were given. Some conclusions were drawn in this paper as follows:(1) With the method of digital image processing and inflexion extracting in the issue, the features of body condition in living pig images were abstracted. These features were correlative with pork grading.(2) The models between manual date and image features were built.The model between pork grading and image features was Y=-1.7456 + 1.3477*x1 + 1.3809*x2 +1.4614* x3 (p<0.05) (Y was pork grade, x1 was slaughter percentage, x2 was lean meat percentage, x3 was back fat thickness)It is feasible that the pig was detected without damage before it was slaughtered with the images. According to the results analyzed in statistics, there are high relativities between back fat thickness, slaughter percentage and lean meat percentage, grades and the feature extracted in image(3) The shape feature parameters in image of the living pig were taken as the inputs of the neural network and then be trained in the NN model. The grades were the outputs of the NN. Its accuracy could reach 92.1%.(4) The detection system of grades of the living pig without damage was established with a good stability. Its accuracy could reach 90.7%.(5 ) The pig can be graded before it is slaughtered by using this system so objectively that the relationship between the pig providers and the purchasers would be improved. The enterprise will get munch economical benefit from the use of this system.