Grading Meat Quality by Image Processing

We study for implementation of a meat quality grading system, using the concept of a "marbling score", as well as image processing, neural network techniques and multiple regression analysis. The marbling score is a measure of the distribution density of fat in the rib-eye region, which is a dominant determinant of the grade of meat. We investigate the aspects of the grading work by graders, and identify five features used for grading in a meat image. For evaluation of the five features, the fat region should be correctly separated from the muscle in the rib-eye region in meat images to be examined. We propose a method of binarization of the image by use of a three-layer neural net trained by a professional grader and a system of meat quality grading based on evaluation of two of five features (the area ratio and the number of large marblings of fat) and multiple regression analysis. Experimental results show that the system is useful.