A nonintrusive beef grading evaluation based on marbling fat imaging by using K-Means Clustering Technique

Abstract

An alternative technique of beef quality assessment and grading was proposed using image analysis based on the K-means clustering algorithm as a semi-automated and nonintrusive approach. An input image of the rib-eye meat area was converted into Lab* color space, followed by classifying groups of features over the spatial domain. By performing a search and cluster, iterating the features of the target image in ab axes resulted in several distinct image zones categorized by different K-groups of features. Marbling fat was estimated in conjunction with quality assessment as an arbitrary grading system. Experimental testing on both synthetic and real image data suggested accuracy better than 95% of averaged. As comparison to conventional methods, this technique is nonintrusive, nondestructive, fast and economic for application and implementation as a measuring tool in the meat research industry.