Prediction of lamb carcass grades using features extracted from lamb chop images Original Research Article

This paper investigates the effectiveness of geometric and texture features extracted from lamb chop images in predicting lamb carcass grade. Twelve geometric and 90 texture (co-occurrence) features were extracted from each of the acquired images. Six feature sets were generated based on the results of dimensionality reduction. These features comprised of three sets of principal component (PC) scores and three sets of reduced features. All six feature sets were used for classification. From the experimental results, it was established that the system enabled 66.3% and 76.9% overall classification based on six PC scores (geometric) and 14 PC scores (geometric and texture), respectively. The system also enabled 64.4% and 79.4% overall classification of lamb carcasses based on six geometric and 14 (geometric and texture) reduced features, respectively. This study shows the predictive potential of combining image analysis with texture analysis for lamb grade prediction. The addition of carcass weight increased the overall classification accuracy, of both feature sets, to 85%.