Combination of Best First and F-Score Image Feature Selection Methods for Beef Freshness Classification

Identification of beef freshness may be conducted by using image processing. The assessment method is done based on image feature extraction. The extracted feature must then be selected only relevant feature in order to obtain an accurate result with an efficient computational load. In this paper, we will compare the F-score and best first methods combined with the F-score. The Best first method is used to select the best features from all the extracted features, then the F-Score method determines the ranking of the best features. The K-Nearest Neighbour (KNN) is used for the classification process. The test is performed by analyzing the results of the classification as values of sensitivity, specificity and accuracy. The results show that performance of meat quality classification has the highest sensitivity, specificity, and accuracy by selecting the combination of Best First and F-Score method. The highest specificity is 0.93 with a sensitivity of 0.8 and an accuracy of 86%. The highest sensitivity, specificity, and accuracy were obtained from the classification by feature selection with a feature subset of the percent mean color intensity R and standard deviation.