Near-infrared hyperspectral imaging for beef tenderness prediction and visualization

Abstract

This research studies rapid and non-destructive prediction of tenderness in fresh beef steaks using near-infrared hyperspectral imaging. Hyperspectral images of tested steaks with different tenderness levels were acquired and their spectral features were extracted in 1000-2500 nm. Calibration algorithms, such as partial least squares regression (PLSR) and least-square support vector machine (LS-SVM) analysis, will be used to correlate the extracted spectra of beef samples with the reference tenderness values estimated by WBSF method. Optimal wavelength selection will be carried out based on full range spectra with two methods, regression coefficients (RC) from PLSR analysis and successful projections algorithm (SPA). An image processing algorithm will be studied to transfer the best model to each pixel in beef steaks for visualizing their WBSF distribution. (AU)