COMPUTER VISION SEGMENTATION OF THE LONGISSIMUS DORSI FOR BEEF QUALITY GRADING

A computer vision system was developed to support automation of beef quality grading. Images of beef steaks were acquired for algorithm development. Fat and lean were differentiated using a fuzzy c-means clustering algorithm. Segmentation of the longissimus dorsi (l.d.) muscle is required because experts assign quality grades based primarily on visual appraisal of the l.d. A robust segmentation algorithm was developed using convex hull procedures. The l.d. was segmented from the steak using morphological operations of erosion and dilation. At the end of each iteration of erosion and dilation, a convex hull was fitted to the image, and compactness was measured. Iterations were continued to yield the most compact l.d. Classification error in segmentation was 1.97%. Average error pixel distance of segmentation by the computer vision system was 4.4 pixels.