Application of Visible Hyperspectral Imaging for Prediction of Springiness of Fresh Chicken Meat

Abstract

Springiness is an important quality characteristic of chicken meat, related with muscle structure and contents of biochemical components, and has great influence on eating quality of chicken meat. Traditional methods for springiness evaluation including manual inspection and instrumental measurement are tedious, time-consuming, and destructive. This study implemented a smart and promising nondestructive technique, i.e., hyperspectral imaging, for rapid prediction of springiness of fresh chicken meat. Hyperspectral images of tested samples with different springiness levels were acquired, and their spectral data were extracted in the spectral range of 400–1,000 nm. Two calibration methods, namely, partial least squares regression (PLSR) and artificial neural network (ANN), were respectively used to correlate the extracted spectra of chicken meat samples with the reference springiness values estimated by a twice-compression method. Successful projections algorithm (SPA) as a popular wavelength selection tool was applied, and ten optimal wavelengths (416, 458, 581, 637, 696, 722, 740, 754, 773, and 973 nm) were finally selected. Based on the selected optimal wavelengths, optimized SPA-PLSR and SPA-ANN model were established, respectively. By comparing with the results of two optimized models, the SPA-PLSR model showed better prediction results with high correlation coefficient (R p) of 0.84 and low root mean square error by prediction (RMSEP) of 0.159. Finally, an image processing algorithm was developed to transfer the SPA-PLSR model to each pixel in chicken meat for visualizing their springiness distribution. The results from the current study indicated that hyperspectral imaging could be a rapid and nondestructive tool for prediction of springiness of chicken meat.