Texture Feature Extraction Based On GLCM and DWT for Beef Tenderness Classification

Four test parameters of beef quality are meat color, fat color, marbling, and tenderness. Beef tenderness measurement's parameters are not workable for visual assessment, because it has no visual standard. One of the tenderness assessment methods requires international standard tools such as Warner-Bratzler. This manual technique has a disadvantage that is not real time and the measurement method causes the beef is not worth consuming because it must be cooked and cut into pieces. The vision parameters that are closely related to tenderness are beef texture. A method has been developed to extract the beef texture as the basic features in the beef classification process. Texture feature's extraction uses statistical analysis of Gray Level Co-occurrence Matrix (GLCM) and domain frequencies Discrete Wavelet Transform (DWT). The correlation test shows that contrast, correlation, energy, and homogeneity features of the GLCM and the diagonal and vertical details of the DWT have significant correlations. The accuracy, precision, and recall for the classification performance reach a hundred percent using Support Vector Machine method with polynomial kernel.