Rapid and nondestructive evaluation of fish freshness by near infrared reflectance spectroscopy combined with chemometrics analysis

Abstract

Rapid and nondestructive measurement of freshness is essential for the control of fish and the quality and safety of its products. In this study, K value was measured by high performance liquid chromatography (HPLC) and employed as an index of fish freshness. The prediction models of the silver chub freshness were developed using Fourier Transform Near Infrared Reflectance Spectroscopy (FT-NIRS) with Several Partial Least Squares (PLS, i-PLS, Si-PLS), Support Vector Machines Regression (SVMR) and Synergy interval plus Support Vector Machine Regression leading to Si-SVMR. By comparison, the performance of Si-SVMR model was superior to the others for the prediction of K value, where RMSECV = 0.027095 and R_c = 95.59% for the calibration set, whereas RMSEP = 0.036525 and R_p = 93.74% for the prediction set. The results indicated that FT-NIR spectroscopy together with Si-SVMR model could be a reliable method for the detection of fish freshness.