Grading of Beef Marbling Based on Image Processing and Support Vector Machine

Beef marbling is the most important indicator of beef quality grading via measuring the abundance of intramuscular fat (IMF) in rib-eye muscle. A beef marbling grading method was developed herein based on image processing and support vector machine (SVM). 123 images of beef rib eye steak were acquired for manual evaluation and image processing. After the marbling, scores were labelled to each image by 5 expert graders; several steps of image processing algorithm were used to extract marbling features, boundary tracking operation for background removal, Otsu's thresholding for fat segmentation, morphological operation and logical operation. Seven features computed from the processed images were used as the input for SVM classifier. The optimum SVM classifier was chosen according to the maximum accuracy of K-fold cross validation based on the data of training set, and then was validated by an independent test set. The accurate rate of the proposed method at 86.0465% shows that the image processing technology combined with SVM algorithm can effectively predict beef marbling scores.