Non-destructive prediction and visualization of chemical composition in lamb meat using NIR hyperspectral imaging and multivariate regression

Hyperspectral reflectance imaging technique for visualization of moisture distribution in cooked chicken breast

Spectroscopy has proven to be an efficient tool for measuring the properties of meat. In this article, hyperspectral imaging (HSI) techniques are used to determine the moisture content in cooked chicken breast over the VIS/NIR (400-1,000 nm) spectral range. Moisture measurements were performed using an oven drying method. A partial least squares regression (PLSR) model was developed to extract a relationship between the HSI spectra and the moisture content. In the full wavelength range, the PLSR model possessed a maximum  of 0.90 and an SEP of 0.74%. For the NIR range, the PLSR model yielded an  of 0.94 and an SEP of 0.71%. The majority of the absorption peaks occurred around 760 and 970 nm, representing the water content in the samples. Finally, PLSR images were constructed to visualize the dehydration and water distribution within different sample regions. The high correlation coefficient and low prediction error from the PLSR analysis validates that HSI is an effective tool for visualizing the chemical properties of meat.

Nondestructive measurement of total volatile basic nitrogen (TVB-N) in pork meat by integrating near infrared spectroscopy, computer vision and electronic nose techniques

Total volatile basic nitrogen (TVB-N) content is an important reference index for evaluating pork freshness. This paper attempted to measure TVB-N content in pork meat using integrating near infrared spectroscopy (NIRS), computer vision (CV), and electronic nose (E-nose) techniques. In the experiment, 90 pork samples with different freshness were collected for data acquisition by three different techniques, respectively. Then, the individual characteristic variables were extracted from each sensor. Next, principal component analysis (PCA) was used to achieve data fusion based on these characteristic variables from 3 different sensors data. Back-propagation artificial neural network (BP-ANN) was used to construct the model for TVB-N content prediction, and the top principal components (PCs) were extracted as the input of model. The result of the model was achieved as follows: the root mean square error of prediction (RMSEP) = 2.73 mg/100g and the determination coefficient (R(p)(2)) = 0.9527 in the prediction set. Compared with single technique, integrating three techniques, in this paper, has its own superiority. This work demonstrates that it has the potential in nondestructive detection of TVB-N content in pork meat using integrating NIRS, CV and E-nose, and data fusion from multi-technique could significantly improve TVB-N prediction performance.