Non-invasive quality analysis of thawed tuna using near infrared spectroscopy with baseline correction

摘要

In order to investigate the ability of NIR spectroscopy coupled with baseline correction to detect quality-related changes of thawed tuna during the storage and the transportation. First, adaptive iteratively reweighted penalized least squares (airPLS) method was taken to eliminate the baseline and estimate spectra feature peaks of 168 samples (bigeye = 82, yellowfin = 86), and the Monte Carlo method was used to determine the optimal adjustment parameter lambda. Second, predictive models based on the fitted features were established by the partial least squares regression (PLSR) to estimate docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA), fat, and protein content. Compared with the standard normal variate (SNV), the proposed airPLS achieved good performance in a comparative experimental test. When the original spectra was preprocessed by airPLS, the number of feature peaks (n(f)) and the mean value ofR(2)(R over bar 2) achieved 6 and 0.89 respectively, which is greater than that of the preprocessing algorithm SNV (n(f)= 4andR over bar 2=0.79). The proposed airPLS effectively reduces the interfering variability and reserves much more NIR features, which is a feasible and stable method for quality-related analysis of the thawed tuna.