The aim here was to explore the potential of visible and near-infrared (Vis/NIR) hyperspectral imaging (400–1000 nm) to classify fresh chicken breast fillets into different water-holding capacity (WHC) groups. Initially, the extracted spectra and image textural features, as well as the mixed data of the two, were used to develop partial least square-discriminant analysis (PLS-DA) classification models. Smoothing, a first derivative process, and principle component analysis (PCA) were carried out sequentially on the mean spectra of all samples to deal with baseline offsets and identify outlier data. Six samples located outside the confidence ellipses of 95% confidence level in the score plot were defined as outliers. A PLS-DA model based on the outlier-free spectra provided a correct classification rate (CCR) value of 78% in the prediction set. Then, seven optimal wavelengths selected using a successive projections algorithm (SPA) were used to develop a simplified PLS-DA model that obtained a slightly reduced CCR with a value of 73%. Moreover, the gray-level co-occurrence matrix (GLCM) was implemented on the first principle component image (with 98.13% of variance) of the hyperspectral image to extract textural features (contrast, correlation, energy, and homogeneity). The CCR of the model developed using textural variables was less optimistic with a value of 59%. Compared to results of models based on spectral or textural data individually, the performance of the model based on the mixed data of optimal spectral and textural features was the best with an improved CCR of 86%. The results showed that the spectral and textural data of hyperspectral images together can be integrated in order to measure and classify the WHC of fresh chicken breast fillets.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited