Prediction of beef chemical composition by NIR Hyperspectral ... - ATB

103104105106107108109110111112113114115116117118119120121122123124125126127FIGURE 1 (a) Assortment of some NIR sub-images at wavelengths indicated, (b) Spectral3.2. Prediction of chemical constituentsfeatures of the examined muscles.Table 1 shows the main statistical parameters resulted from each model in cross validationand prediction samples. The results indicated that the PLSR models were very efficient inpredicting these chemical constituents with a determination coefficient (2Rcv) of 0.91, 0.89and 0.91 using 9, 8, 10 factors for fat, moisture and protein respectively. When the modelused in predicting these constituents in the samples of a prediction set, the model gave goodprediction ability with determination coefficient (2Rp) of 0.89, 0.86 and 0.75 for water, fat andprotein, respectively. However, the predictability of protein content might be considereduncertain as the value of2RP(0.75) is rather small and the number of latent factors washigher compared with the other two constituents. The lack of a strong predictability forprotein in meat suing spectral analyses is not new, and several authors have reported similarresults (Alomar 2003).It was very advantageous to select significant variables during a multivariate regression inorder to improve the predictive ability of the model and to augment the processing speed.These feature-related wavelengths identified from the PLSR’s weighted regressioncoefficients of each constituent indicated that eight wavelengths (934, 1048, 1108, 1155,1185, 1212, 1265, 1379 nm) were defined as the most relevant wavelengths in predictingwater contents. Similarly, seven wavelengths (934, 978, 1078, 1138, 1215, 1289, 1413 nm)were selected for fat and ten wavelengths (924, 937, 1018, 1048, 1108, 1141, 1182, 1221,1615, and 1665 nm) were selected for efficient prediction of protein. The results shown inTable 1 revealed that the predictability of these models is still good indicating the robustness